HIBPool.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "HIBPool.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "include_colab_link": true
    },
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.7.8"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/forkkr/HIBPool/blob/main/HIBPool.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nxaWWxB4gE_x"
      },
      "source": [
        "#Description\n",
        "DiffPool with Louvain based community\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kzPlMnhngITT"
      },
      "source": [
        "#Install Packages"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iDelXzLYgJMj",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "924134ac-b842-44c4-9247-194f80ac0c83"
      },
      "source": [
        "!pip install utils\n",
        "!pip install pyhocon\n",
        "!pip install scikit-learn-extra\n",
        "!pip install tensorboardX\n",
        "!pip install dgl==0.4\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting utils\n",
            "  Downloading https://files.pythonhosted.org/packages/55/e6/c2d2b2703e7debc8b501caae0e6f7ead148fd0faa3c8131292a599930029/utils-1.0.1-py2.py3-none-any.whl\n",
            "Installing collected packages: utils\n",
            "Successfully installed utils-1.0.1\n",
            "Collecting pyhocon\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/af/c4/39cfc164ed77ddfe7dd80ff31b065c2b18583c4ee0a4336979c86c4b5d8f/pyhocon-0.3.57.tar.gz (110kB)\n",
            "\u001b[K     |████████████████████████████████| 112kB 5.1MB/s \n",
            "\u001b[?25hRequirement already satisfied: pyparsing>=2.0.3 in /usr/local/lib/python3.6/dist-packages (from pyhocon) (2.4.7)\n",
            "Building wheels for collected packages: pyhocon\n",
            "  Building wheel for pyhocon (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for pyhocon: filename=pyhocon-0.3.57-cp36-none-any.whl size=18540 sha256=fb059b2e7cd50929ed16fe178b3e3170cd0665c12d135b56778ff324f517e77b\n",
            "  Stored in directory: /root/.cache/pip/wheels/43/ae/39/250b80e3aa8082c5fbb3f14782b9cad47de0b885f22e927a8f\n",
            "Successfully built pyhocon\n",
            "Installing collected packages: pyhocon\n",
            "Successfully installed pyhocon-0.3.57\n",
            "Collecting scikit-learn-extra\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/d1/dd/891d2ee7bd18af8f2e1df5d63a52ee96edd0eacc21bb9627072b1c5f6a6c/scikit-learn-extra-0.1.0b2.tar.gz (615kB)\n",
            "\u001b[K     |████████████████████████████████| 624kB 5.5MB/s \n",
            "\u001b[?25h  Installing build dependencies ... \u001b[?25l\u001b[?25hdone\n",
            "  Getting requirements to build wheel ... \u001b[?25l\u001b[?25hdone\n",
            "    Preparing wheel metadata ... \u001b[?25l\u001b[?25hdone\n",
            "Requirement already satisfied: scikit-learn>=0.21.0 in /usr/local/lib/python3.6/dist-packages (from scikit-learn-extra) (0.22.2.post1)\n",
            "Requirement already satisfied: numpy>=1.13.3 in /usr/local/lib/python3.6/dist-packages (from scikit-learn-extra) (1.19.5)\n",
            "Requirement already satisfied: scipy>=0.19.1 in /usr/local/lib/python3.6/dist-packages (from scikit-learn-extra) (1.4.1)\n",
            "Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.6/dist-packages (from scikit-learn>=0.21.0->scikit-learn-extra) (1.0.0)\n",
            "Building wheels for collected packages: scikit-learn-extra\n",
            "  Building wheel for scikit-learn-extra (PEP 517) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for scikit-learn-extra: filename=scikit_learn_extra-0.1.0b2-cp36-cp36m-linux_x86_64.whl size=339581 sha256=6489d653cdcec10538abf8cdb256bba07df2bc96ba6cfb3f1f0b6c2e00eb44a7\n",
            "  Stored in directory: /root/.cache/pip/wheels/04/01/0f/943bffb48bac048fa216b4325f1a6c939491ccb0ff500e08f4\n",
            "Successfully built scikit-learn-extra\n",
            "Installing collected packages: scikit-learn-extra\n",
            "Successfully installed scikit-learn-extra-0.1.0b2\n",
            "Collecting tensorboardX\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/af/0c/4f41bcd45db376e6fe5c619c01100e9b7531c55791b7244815bac6eac32c/tensorboardX-2.1-py2.py3-none-any.whl (308kB)\n",
            "\u001b[K     |████████████████████████████████| 317kB 4.1MB/s \n",
            "\u001b[?25hRequirement already satisfied: six in /usr/local/lib/python3.6/dist-packages (from tensorboardX) (1.15.0)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from tensorboardX) (1.19.5)\n",
            "Requirement already satisfied: protobuf>=3.8.0 in /usr/local/lib/python3.6/dist-packages (from tensorboardX) (3.12.4)\n",
            "Requirement already satisfied: setuptools in /usr/local/lib/python3.6/dist-packages (from protobuf>=3.8.0->tensorboardX) (51.1.2)\n",
            "Installing collected packages: tensorboardX\n",
            "Successfully installed tensorboardX-2.1\n",
            "Collecting dgl==0.4\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/26/66/62c506d1d9b754cf36a6cee4eccc51c517d3b49ec85cc60ff48b04d7e23b/dgl-0.4-cp36-cp36m-manylinux1_x86_64.whl (2.3MB)\n",
            "\u001b[K     |████████████████████████████████| 2.3MB 4.3MB/s \n",
            "\u001b[?25hRequirement already satisfied: scipy>=1.1.0 in /usr/local/lib/python3.6/dist-packages (from dgl==0.4) (1.4.1)\n",
            "Requirement already satisfied: networkx>=2.1 in /usr/local/lib/python3.6/dist-packages (from dgl==0.4) (2.5)\n",
            "Requirement already satisfied: numpy>=1.14.0 in /usr/local/lib/python3.6/dist-packages (from dgl==0.4) (1.19.5)\n",
            "Requirement already satisfied: decorator>=4.3.0 in /usr/local/lib/python3.6/dist-packages (from networkx>=2.1->dgl==0.4) (4.4.2)\n",
            "Installing collected packages: dgl\n",
            "Successfully installed dgl-0.4\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2g4bEDHYgLWs"
      },
      "source": [
        "#Import Packages"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZWDY-4vbgDm-"
      },
      "source": [
        "\n",
        "# Default\n",
        "\n",
        "import os\n",
        "import re\n",
        "import sys\n",
        "import math\n",
        "import random\n",
        "import pyhocon\n",
        "import argparse\n",
        "import itertools\n",
        "import seaborn\n",
        "import atexit\n",
        "import shutil\n",
        "from tqdm import tqdm\n",
        "import utils\n",
        "import time\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "from collections import defaultdict\n",
        "\n",
        "from __future__ import division\n",
        "from __future__ import print_function\n",
        "# from tensorboardX import SummaryWriter\n",
        "\n",
        "import dgl\n",
        "from dgl.data.utils import download, extract_archive, get_download_dir, _get_dgl_url\n",
        "\n",
        "\n",
        "# PyTorch\n",
        "\n",
        "import torch\n",
        "import numpy as np\n",
        "import scipy\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "from torch.autograd import Variable\n",
        "import torch.optim as optim\n",
        "import torch.utils.data\n",
        "from torch.utils.data import Dataset,DataLoader, random_split\n",
        "\n",
        "# Sklearn\n",
        "\n",
        "from sklearn.utils import shuffle\n",
        "from sklearn.metrics import f1_score\n",
        "from sklearn.metrics import accuracy_score\n",
        "\n",
        "# Utility\n",
        "import community as community_louvain\n",
        "import networkx as nx\n",
        "from sklearn.manifold import TSNE\n",
        "import seaborn as sns\n",
        "import matplotlib.patheffects as PathEffects\n",
        "import time\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "obuocLRom1hL"
      },
      "source": [
        "#Utility"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bCtMSPvzNea7"
      },
      "source": [
        "## Import GIB Utility"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "A0Mx4BqQM5h2"
      },
      "source": [
        "sys.path.append(os.path.abspath('/content/drive/MyDrive/Graph_Pooling'))\n",
        "from GIB import *"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1wpnMeJxutui"
      },
      "source": [
        "## Gaussian XIB"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "u6uHlMj0JVEa"
      },
      "source": [
        "from torch.distributions.normal import Normal\n",
        "def gaussian_xib(embs, s_l, dim, device):\n",
        "  assert embs.shape[1] == dim\n",
        "  dim = int(dim/2)\n",
        "  prior = Normal(loc=torch.zeros(s_l.shape[1], dim).to(device),\n",
        "                 scale=torch.ones(s_l.shape[1], dim).to(device),)\n",
        "  posterior, _ = reparameterize(model=None, input=embs, mode='diag', size=dim)\n",
        "\n",
        "  # print(posterior.shape, s_l.shape)\n",
        "  # n1, n2 = torch.nonzero(s_l, as_tuple=True)\n",
        "  # posterior = posterior[n2]\n",
        "  # print(posterior.shape, s_l.shape)\n",
        "  # print(prior.shape)\n",
        "  # print('n1: {} n2: {}'.format(n1, n2))\n",
        "  xib = torch.distributions.kl.kl_divergence(posterior, prior).sum(-1).unsqueeze(0).mean(1).unsqueeze(0)\n",
        "  embs = embs[:, :dim]\n",
        "  # print(xib.shape, embs.shape)\n",
        "  return embs, xib"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aMcwgHbfIR3D"
      },
      "source": [
        "##Feature Perturbation"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7nxJ_mtHIYhV"
      },
      "source": [
        "\n",
        "def feature_perturb(x, perturb_ratio):\n",
        "  x_max_mean = 0\n",
        "  x_c = []\n",
        "  if perturb_ratio != 0.0:\n",
        "    for tx in x:\n",
        "      # print(tx)\n",
        "      # print(type(tx), tx.shape)\n",
        "      # print(tx.max(1)[0].shape)\n",
        "      tx_c = tx + torch.randn(tx.shape) * tx.max(1)[0].mean() * perturb_ratio\n",
        "      # x + torch.randn(x.shape) * x_max_mean * perturb_ratio\n",
        "      x_c.append(tx_c)\n",
        "  else:\n",
        "    x_c = deepcopy(x)\n",
        "  \n",
        "  return x_c\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "e6yYnSeD6CL3"
      },
      "source": [
        "#Data Loader"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sxgRDNVw6Eby"
      },
      "source": [
        "class TUDataset(Dataset):\n",
        "    _url = r\"https://ls11-www.cs.uni-dortmund.de/people/morris/graphkerneldatasets/{}.zip\"\n",
        "\n",
        "    def __init__(self, name):\n",
        "\n",
        "        self.name = name\n",
        "\n",
        "        # self.extract_dir = self._download()\n",
        "        self.extract_dir = '/content/drive/My Drive/Graph_Pooling'\n",
        "#         self.extract_dir = os.getcwd() + '/data'\n",
        "        DS_edge_list = self._idx_from_zero(\n",
        "            np.loadtxt(self._file_path(\"A\"), delimiter=\",\", dtype=int))\n",
        "        DS_indicator = self._idx_from_zero(\n",
        "            np.loadtxt(self._file_path(\"graph_indicator\"), dtype=int))\n",
        "        DS_graph_labels = self._idx_from_zero(\n",
        "            np.loadtxt(self._file_path(\"graph_labels\"), dtype=int))\n",
        "\n",
        "        g = dgl.DGLGraph()\n",
        "        g.add_nodes(int(DS_edge_list.max()) + 1)\n",
        "        g.add_edges(DS_edge_list[:, 0], DS_edge_list[:, 1])\n",
        "\n",
        "        node_idx_list = []\n",
        "        for idx in range(np.max(DS_indicator) + 1):\n",
        "            node_idx = np.where(DS_indicator == idx)\n",
        "            node_idx_list.append(node_idx[0])\n",
        "        self.graph_lists = g.subgraphs(node_idx_list)\n",
        "        self.graph_labels = DS_graph_labels\n",
        "\n",
        "        try:\n",
        "            DS_node_labels = self.to_onehot(\n",
        "                self._idx_from_zero(np.loadtxt(self._file_path(\"node_labels\"), dtype=int)))\n",
        "            for idxs, g in zip(node_idx_list, self.graph_lists):\n",
        "                g.ndata['feat'] = DS_node_labels[idxs, :]\n",
        "        except IOError:\n",
        "            print(\"No Node Label Data\")\n",
        "\n",
        "        try:\n",
        "            DS_node_attr = np.loadtxt(self._file_path(\"node_attributes\"), delimiter=\",\")\n",
        "            shape2 = DS_node_attr.shape[1] if len(DS_node_attr.shape)>1 else 1\n",
        "            DS_node_attr = DS_node_attr.reshape(DS_node_attr.shape[0],shape2)\n",
        "            for idxs, g in zip(node_idx_list, self.graph_lists):\n",
        "                g.ndata['feat'] = DS_node_attr[idxs,:]\n",
        "        except IOError:\n",
        "            print(\"No Node Attribute Data\")\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        g = self.graph_lists[idx]\n",
        "        return (g.adjacency_matrix().to_dense(), g.ndata['feat']), self.graph_labels[idx]\n",
        "\n",
        "    def __len__(self):\n",
        "        return len(self.graph_lists)\n",
        "\n",
        "    def _download(self):\n",
        "        download_dir = get_download_dir()\n",
        "        zip_file_path = os.path.join(download_dir, \"tu_{}.zip\".format(self.name))\n",
        "        download(self._url.format(self.name), path=zip_file_path)\n",
        "        extract_dir = os.path.join(download_dir, \"tu_{}\".format(self.name))\n",
        "        extract_archive(zip_file_path, extract_dir)\n",
        "        return extract_dir\n",
        "\n",
        "    def _file_path(self, category):\n",
        "        return os.path.join(self.extract_dir, self.name, \"{}_{}.txt\".format(self.name, category))\n",
        "\n",
        "    @staticmethod\n",
        "    def _idx_from_zero(idx_tensor):\n",
        "        return idx_tensor - np.min(idx_tensor)\n",
        "\n",
        "    @staticmethod\n",
        "    def to_onehot(label_tensor):\n",
        "        label_num = label_tensor.shape[0]\n",
        "        assert np.min(label_tensor) == 0\n",
        "        one_hot_tensor = np.zeros((label_num, np.max(label_tensor) + 1))\n",
        "        one_hot_tensor[np.arange(label_num), label_tensor] = 1\n",
        "        return one_hot_tensor\n",
        "\n",
        "\n",
        "class CollateFn:\n",
        "    def __init__(self, device='cpu'):\n",
        "        self.device = device\n",
        "\n",
        "    def __call__(self, batch):\n",
        "        adj_tensor_list = []\n",
        "        features_list = []\n",
        "        mask_list = []\n",
        "        node_list = []\n",
        "        # (adj, features), labels = list(zip(*batch))\n",
        "        max_num_nodes = max([g[0][0].shape[0] for g in batch])\n",
        "        labels = []\n",
        "        for (A, F), L in batch:\n",
        "            adj_tensor_list.append(A.float())\n",
        "            features_list.append(F.float())\n",
        "            labels.append(L)\n",
        "            # labels.append(L)\n",
        "            # length = A.shape[0]\n",
        "            # pad_len = max_num_nodes - length  #up, down,   left, right\n",
        "            # adj_tensor_list.append(np.pad(A, ((0, pad_len), (0, pad_len)), mode='constant'))\n",
        "            # features_list.append(np.pad(F, ((0, pad_len), (0, 0)), mode='constant'))\n",
        "            # node_list.append(length)\n",
        "            # mask = np.zeros(max_num_nodes)\n",
        "            # mask[:length] = 1\n",
        "            # mask_list.append(mask)\n",
        "        # return torch.from_numpy(np.stack(adj_tensor_list, 0)).float().to(self.device), \\\n",
        "        #        torch.from_numpy(np.stack(features_list, 0)).float().to(self.device), \\\n",
        "        #        torch.from_numpy(np.stack(mask_list, 0)).float().to(self.device), \\\n",
        "        #        torch.from_numpy(np.stack(labels, 0)).long().to(self.device), \\\n",
        "        #        torch.from_numpy(np.stack(node_list, 0)).long().to(self.device)\n",
        "        return adj_tensor_list,features_list,\\\n",
        "               torch.from_numpy(np.stack(labels, 0)).long().to(self.device)\n",
        "\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EuVtx1PQ52J3"
      },
      "source": [
        "#Models"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3qlcMnvJ5jTc"
      },
      "source": [
        "class BatchedGraphSAGE(nn.Module):\n",
        "    def __init__(self, infeat, outfeat, device='cpu', use_bn=True, mean=False, add_self=False):\n",
        "        super().__init__()\n",
        "        self.add_self = add_self\n",
        "        self.use_bn = use_bn\n",
        "        self.device = device\n",
        "        self.mean = mean\n",
        "        self.W = nn.Linear(2*infeat, outfeat, bias=True)\n",
        "        # nn.init.xavier_uniform_(self.W.weight, gain=nn.init.calculate_gain('relu'))\n",
        "\n",
        "    def forward(self, x, adj, mask=None):\n",
        "        # print(type(adj), type(x))\n",
        "        if self.add_self:\n",
        "            adj = adj \n",
        "            # + torch.eye(adj.size(0)).to(self.device)\n",
        "\n",
        "        if self.mean:\n",
        "            adj = adj / adj.sum(1, keepdim=True)\n",
        "\n",
        "        h_k_N = torch.matmul(adj, x)\n",
        "        h_k = self.W(torch.cat([x,h_k_N], dim=1))\n",
        "        h_k = F.normalize(h_k, dim=1, p=2)\n",
        "        h_k = F.relu(h_k)\n",
        "        if self.use_bn and h_k.shape[0]>1:\n",
        "            self.bn = nn.BatchNorm1d(h_k.size(0)).to(self.device)\n",
        "            h_k = self.bn(h_k)\n",
        "        if mask is not None:\n",
        "            h_k = h_k * mask.unsqueeze(2).expand_as(h_k)\n",
        "        return h_k\n",
        "\n",
        "\n",
        "class BatchedDiffPool(nn.Module):\n",
        "    def __init__(self, nfeat, nhid, is_final=False, device='cpu', link_pred=False):\n",
        "        super(BatchedDiffPool, self).__init__()\n",
        "        self.link_pred = link_pred\n",
        "        self.device = device\n",
        "        self.is_final = is_final\n",
        "        double_nhid = nhid*2 \n",
        "        self.mlp1 = nn.Linear(nfeat, nhid, bias=True)\n",
        "        self.mlp2 = nn.Linear(nhid, double_nhid, bias=True)\n",
        "        self.agg_param = nn.Parameter(torch.FloatTensor(12*double_nhid,double_nhid))\n",
        "        self.mlp3 = nn.Linear(double_nhid, double_nhid, bias=True)\n",
        "        self.mlp4 = nn.Linear(double_nhid, double_nhid, bias=True)\n",
        "        self.MPN = BatchedGraphSAGE(nhid, nhid, device=self.device, use_bn=False)\n",
        "        self.log = {}\n",
        "        self.link_pred_loss = 0\n",
        "        self.entropy_loss = 0\n",
        "\n",
        "        self.out_channels = nhid\n",
        "        self.nhid = double_nhid\n",
        "        self.init_params()\n",
        "        # self.device = device\n",
        "      \n",
        "    \n",
        "    def init_params(self):\n",
        "      for param in self.parameters():\n",
        "        if len(param.size()) == 2:\n",
        "          nn.init.xavier_uniform_(param,gain=nn.init.calculate_gain('relu'))\n",
        "\n",
        "\n",
        "    \n",
        "    def get_assign_mat(self,adj):\n",
        "      \n",
        "      \n",
        "      n1 = adj.shape[0]\n",
        "\n",
        "      # rows, cols = np.where(adj == 1)\n",
        "      rows, cols = torch.nonzero(adj, as_tuple = True)\n",
        "      edges = zip(rows.tolist(), cols.tolist())\n",
        "      \n",
        "      G = nx.Graph()\n",
        "      for u in range(n1):\n",
        "        G.add_node(u)\n",
        "      G.add_edges_from(edges)\n",
        "      cluster_coeff = nx.clustering(G)\n",
        "      deg_centrality = nx.degree_centrality(G)\n",
        "      \n",
        "\n",
        "      dendo = community_louvain.generate_dendrogram(G)\n",
        "      assign_mat_list = []\n",
        "      cc_mat_list = []\n",
        "      deg_mat_list = []\n",
        "      # print(len(dendrogram))\n",
        "      # node2com = community_louvain.best_partition(G)\n",
        "\n",
        "      for level in range(len(dendo)-1,max(-1,len(dendo)-3),-1):\n",
        "        node2com = community_louvain.partition_at_level(dendo, level)\n",
        "        \n",
        "        com2node = dict()\n",
        "        for node in node2com:\n",
        "          if node2com[node] not in com2node:\n",
        "            com2node[node2com[node]] = list()\n",
        "          com2node[node2com[node]].append(node)\n",
        "        \n",
        "        n2 = len(com2node)\n",
        "        assign_mat = torch.zeros(n1,n2).float().to(self.device)\n",
        "        \n",
        "        for u in com2node:\n",
        "          for v in com2node[u]:\n",
        "            assign_mat[v][u] = 1\n",
        "        \n",
        "        cluster_coeff = torch.exp(torch.FloatTensor([cluster_coeff[u] for u in range(n1)])).reshape(n1, 1).to(self.device)\n",
        "        deg_centrality = torch.exp(torch.FloatTensor([deg_centrality[u] for u in range(n1)])).reshape(n1, 1).to(self.device)\n",
        "        \n",
        "        cls_sum = torch.mm(torch.transpose(assign_mat, -1, 0), cluster_coeff)\n",
        "        deg_sum = torch.mm(torch.transpose(assign_mat, -1, 0), deg_centrality)\n",
        "\n",
        "        cls_sum = torch.mm(assign_mat, cls_sum)\n",
        "        deg_sum = torch.mm(assign_mat, deg_sum)\n",
        "\n",
        "        cc_mat = cluster_coeff*(1/cls_sum)\n",
        "        deg_mat = deg_centrality*(1/deg_sum)\n",
        "        \n",
        "        assign_mat_list.append(assign_mat)\n",
        "        cc_mat_list.append(cc_mat)\n",
        "        deg_mat_list.append(deg_mat)\n",
        "\n",
        "      return assign_mat_list,cc_mat_list,deg_mat_list\n",
        "\n",
        "    def aggregator(self,z_l,s_l):\n",
        "      n1,n2 = s_l.shape[0],s_l.shape[1]\n",
        "      sum_feat = torch.matmul(s_l.transpose(-1, -2), z_l).to(self.device)\n",
        "      col_sum = torch.sum(s_l,dim=0).reshape(s_l.shape[1],1).to(self.device)\n",
        "      col_sum = torch.where(col_sum>0,col_sum,torch.ones(col_sum.shape, device=self.device))\n",
        "      mean_feat = sum_feat/col_sum\n",
        "\n",
        "      max_feat = []\n",
        "      \n",
        "      for j in range(n2):\n",
        "        temp = torch.zeros(1,self.nhid, device=self.device)\n",
        "        for i in range(n1):\n",
        "          if s_l[i][j] == 1:\n",
        "            cur_feat = z_l[i].reshape(1,self.nhid)\n",
        "            con = torch.cat([temp,cur_feat],dim=0)\n",
        "            temp = torch.max(con,dim=0).values.reshape(1,self.nhid)\n",
        "        max_feat.append(temp)\n",
        "      max_feat = torch.cat(max_feat,dim=0)\n",
        "\n",
        "      min_feat = []\n",
        "      n1,n2 = s_l.shape[0],s_l.shape[1]\n",
        "      for j in range(n2):\n",
        "        temp = []\n",
        "        for i in range(n1):\n",
        "          if s_l[i][j] == 1:\n",
        "            cur_feat = z_l[i].reshape(1,self.nhid)\n",
        "            if len(temp) == 0:\n",
        "              temp = cur_feat\n",
        "            else:\n",
        "              con = torch.cat([temp,cur_feat],dim=0)\n",
        "              temp = torch.min(con,dim=0).values.reshape(1,self.nhid)\n",
        "        if len(temp)== 0:\n",
        "          temp = torch.zeros(1,self.nhid)\n",
        "        min_feat.append(temp)\n",
        "      min_feat = torch.cat(min_feat,dim=0)\n",
        "      return sum_feat, mean_feat, max_feat, min_feat\n",
        "\n",
        "      \n",
        "    def run_Louvain(self,x,adj):\n",
        "      b_sz = len(x)\n",
        "      batch_gxib = []\n",
        "      final_feat = []\n",
        "      for i in range(b_sz):\n",
        "        feat = x[i].to(self.device)\n",
        "        g = adj[i].to(self.device)\n",
        "        \n",
        "        feat = self.mlp1(feat)\n",
        "        feat = self.MPN(feat, g)\n",
        "        feat = F.relu(feat)\n",
        "        feat = self.mlp2(feat)\n",
        "        feat = F.relu(feat)\n",
        "        \n",
        "        feat_original = feat\n",
        "        \n",
        "        assign_mat_list,cc_mat_list,deg_mat_list = self.get_assign_mat(g)\n",
        "        graph_feat = torch.zeros(1,self.out_channels).float().to(self.device)\n",
        "        for i in range(len(assign_mat_list)):\n",
        "\n",
        "            s_l = assign_mat_list[i]\n",
        "            cc_mat = cc_mat_list[i]\n",
        "            deg_mat = deg_mat_list[i]\n",
        "\n",
        "            n1 = s_l.shape[0]\n",
        "            n2 = s_l.shape[1]\n",
        "\n",
        "            z_l = feat_original\n",
        "            feat = []\n",
        "            sum, mean,mx,mn = self.aggregator(z_l,s_l)\n",
        "            feat.append(sum)\n",
        "            feat.append(mean)\n",
        "            feat.append(mx)\n",
        "            feat.append(mn)\n",
        "            # sum, mean,mx,mn = self.aggregator(cc_mat.mm(z_l),s_l)\n",
        "            sum, mean,mx,mn = self.aggregator(cc_mat*z_l,s_l)\n",
        "            feat.append(sum)\n",
        "            feat.append(mean)\n",
        "            feat.append(mx)\n",
        "            feat.append(mn)\n",
        "            # sum, mean, mx, mn = self.aggregator(deg_mat.mm(z_l),s_l)\n",
        "            sum, mean, mx, mn = self.aggregator(deg_mat*z_l,s_l)\n",
        "            feat.append(sum)\n",
        "            feat.append(mean)\n",
        "            feat.append(mx)\n",
        "            feat.append(mn)\n",
        "\n",
        "            feat = torch.cat(feat,dim=0).reshape(n2,self.nhid*12)\n",
        "            feat = feat.mm(self.agg_param)\n",
        "\n",
        "\n",
        "            feat = self.mlp3(feat)\n",
        "            feat = F.relu(feat)\n",
        "            feat = self.mlp4(feat)\n",
        "            feat = F.relu(feat)\n",
        "\n",
        "            # h1, h2 = torch.nonzero(s_l, as_tuple=True)\n",
        "            # ext_embs = feat[h2]\n",
        "            _, gxib = gaussian_xib(feat, s_l, feat.shape[1], self.device)\n",
        "            batch_gxib.append(gxib)\n",
        "\n",
        "            feat = feat[:,:self.out_channels]\n",
        "\n",
        "            feat = torch.sum(feat,dim=0).reshape(1,self.out_channels)\n",
        "            graph_feat += feat\n",
        "            \n",
        "        graph_feat/=len(assign_mat_list)\n",
        "        final_feat.append(graph_feat)\n",
        "\n",
        "      final_feat = torch.cat(final_feat,dim=0)\n",
        "      return final_feat, batch_gxib       \n",
        "\n",
        "\n",
        "    def forward(self, x, adj, mask=None, log=False):\n",
        "        final_embd, batch_gxib = self.run_Louvain(x,adj)\n",
        "        return final_embd, batch_gxib\n",
        "\n",
        "\n",
        "class Classifier(nn.Module):\n",
        "    def __init__(self, n_classes):\n",
        "        super().__init__()\n",
        "        self.classifier = nn.Sequential(nn.Linear(30, 50),\n",
        "                                        nn.ReLU(),\n",
        "                                        nn.Linear(50, n_classes))\n",
        "\n",
        "    def forward(self, x):\n",
        "        return self.classifier(x)\n",
        "\n",
        "\n",
        "class BatchedModel(nn.Module):\n",
        "    def __init__(self, pool_size, device, input_shape, n_classes, link_pred=False):\n",
        "        super().__init__()\n",
        "        self.input_shape = input_shape\n",
        "        self.link_pred = link_pred\n",
        "        self.device = device\n",
        "        self.layers = nn.ModuleList([\n",
        "            # BatchedGraphSAGE(input_shape, 30, device=self.device)\n",
        "            # BatchedGraphSAGE(30, 30, device=self.device),\n",
        "            BatchedDiffPool(input_shape, 30, device=self.device, link_pred=link_pred)#,\n",
        "            # BatchedGraphSAGE(30, 30, device=self.device),\n",
        "            # BatchedGraphSAGE(30, 30, device=self.device),\n",
        "            # BatchedDiffPool(30, 1, 30, is_final=True, device=self.device)\n",
        "        ])\n",
        "        self.classifier = Classifier(n_classes)\n",
        "        # writer.add_text(str(vars(self)))\n",
        "        self.init_params()\n",
        "    \n",
        "    def init_params(self):\n",
        "      for param in self.parameters():\n",
        "        if len(param.size()) == 2:\n",
        "          nn.init.xavier_uniform_(param,gain=nn.init.calculate_gain('relu'))\n",
        "\n",
        "    def forward(self, x, adj):\n",
        "        for layer in self.layers:\n",
        "          \n",
        "          x = layer(x, adj)\n",
        "\n",
        "        # x = x * mask\n",
        "        # readout_x = x.sum(dim=1)\n",
        "        readout_x = x\n",
        "        return readout_x\n",
        "\n",
        "    def loss(self, output, labels):\n",
        "        criterion = nn.CrossEntropyLoss()\n",
        "        loss = criterion(output, labels)\n",
        "        # if self.link_pred:\n",
        "        #     for layer in self.layers:\n",
        "        #         if isinstance(layer, BatchedDiffPool):\n",
        "        #             loss = loss + layer.link_pred_loss.mean() + layer.entropy_loss.mean()\n",
        "\n",
        "        return loss"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-xz8LKhn6k4D"
      },
      "source": [
        "#Training"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eAA-AzQh57LR"
      },
      "source": [
        "def main():\n",
        "#     utils.writer = SummaryWriter()\n",
        "\n",
        "    parser = argparse.ArgumentParser()\n",
        "    parser.add_argument('-f')\n",
        "    parser.add_argument('--no-cuda', action='store_true', default=False,\n",
        "                        help='Disables CUDA training.')\n",
        "    parser.add_argument('--epochs', type=int, default=1000,\n",
        "                        help='Number of epochs to train.')\n",
        "    parser.add_argument('--link-pred', action='store_true', default=False,\n",
        "                        help='Enable Link Prediction Loss')\n",
        "    parser.add_argument('--dataset', default='ENZYMES', help=\"Choose dataset: ENZYMES, DD, NCI1\")\n",
        "    parser.add_argument('--batch-size', default=8, type=int, help=\"Choose dataset: ENZYMES, DD\")\n",
        "    parser.add_argument('--train-ratio', default=0.8, type=float, help=\"Train/Val split ratio\")\n",
        "    parser.add_argument('--pool-ratio', default=0.25, type=float, help=\"Train/Val split ratio\")\n",
        "    parser.add_argument('--beta', default=0.01, type=float, help=\"IB beta\")\n",
        "    parser.add_argument('--cuda', type=bool, default=False)\n",
        "    parser.add_argument('--load_run', type=bool, default=False)\n",
        "    parser.add_argument('--save_model', type=bool, default=True)\n",
        "    \n",
        "    PATH = '/content/drive/My Drive/Graph_Pooling'\n",
        "\n",
        "    \n",
        "\n",
        "\t  \n",
        "    \n",
        "    title = 'Test F1,' \\\n",
        "            'Test F1(0.5),' \\\n",
        "            'Test F1(1.0),' \\\n",
        "            'Test F1(1.5),' \\\n",
        "            'Test F1(2.0),' \\\n",
        "            '\\n'\n",
        "\n",
        "    args = parser.parse_args()\n",
        "#     utils.writer.add_text(\"args\", str(args))\n",
        "    device = \"cuda\" if not args.no_cuda and torch.cuda.is_available() else \"cpu\"\n",
        "    # random.seed(args.seed)\n",
        "\t# np.random.seed(args.seed)\n",
        "\t# torch.manual_seed(args.seed)\n",
        "\t# torch.cuda.manual_seed_all(args.seed)\n",
        "\n",
        "    file_name = PATH + '/' + args.dataset + '_summary' + '.csv'\n",
        "\n",
        "    dataset = TUDataset(args.dataset)\n",
        "    # dataset = MNIST(root=\"~/.torch/data/\", transform=GraphTransform(device), download=True)\n",
        "    dataset_size = len(dataset)\n",
        "    train_size = int(dataset_size * args.train_ratio)\n",
        "    test_size = int((dataset_size - train_size)/2)\n",
        "    val_size = dataset_size - (train_size + test_size)\n",
        "    max_num_nodes = max([item[0][0].shape[0] for item in dataset])\n",
        "    n_classes = int(max([item[1] for item in dataset])) + 1\n",
        "    train_data, val_data, test_data = random_split(dataset, (train_size, val_size, test_size))\n",
        "    input_shape = int(dataset[0][0][1].shape[-1])\n",
        "    train_loader = DataLoader(train_data, batch_size=args.batch_size, shuffle=True,\n",
        "                              collate_fn=CollateFn(device))\n",
        "    test_loader = DataLoader(test_data, batch_size=args.batch_size, shuffle=True,\n",
        "                             collate_fn=CollateFn(device))\n",
        "    val_loader = DataLoader(val_data, batch_size=args.batch_size, shuffle=True,\n",
        "                             collate_fn=CollateFn(device))\n",
        "\n",
        "    model = BatchedModel(pool_size=int(max_num_nodes * args.pool_ratio), device=device,\n",
        "                         link_pred=args.link_pred, input_shape=input_shape, n_classes=n_classes).to(\n",
        "        device)\n",
        "    model.train()\n",
        "    optimizer = optim.Adam(model.parameters())\n",
        "\n",
        "    train_iter = 0\n",
        "    epoch_losses_list = []\n",
        "    ep_vt = []\n",
        "    test_iter = 0\n",
        "    mx_acc = [0,0,0,0,0]\n",
        "    version = 0\n",
        "    for e in range(args.epochs):\n",
        "        print(\"Epoch: \",e)\n",
        "        utils.e = e\n",
        "        \n",
        "        true_sample = 0\n",
        "        model.train()\n",
        "        st = time.time()\n",
        "        for i, (adj, features, batch_labels) in enumerate(train_loader):\n",
        "\n",
        "            train_iter += 1\n",
        "            graph_feat, batch_gxib = model(features, adj)\n",
        "            output = model.classifier(graph_feat)\n",
        "            loss = model.loss(output, batch_labels)\n",
        "\n",
        "            # print(batch_gxib)\n",
        "            # print(torch.cat(batch_gxib, dim=0).shape)\n",
        "            gxib_loss = torch.cat(batch_gxib, dim=0).mean()\n",
        "\n",
        "            loss = loss + args.beta*gxib_loss \n",
        "\n",
        "            loss.backward()\n",
        "            torch.nn.utils.clip_grad_norm_(model.parameters(), 2.0)\n",
        "            optimizer.step()\n",
        "            optimizer.zero_grad()\n",
        "\n",
        "            epoch_losses_list.append(loss.item())\n",
        "            ep_vt.append(train_iter)\n",
        "            iter_true_sample = (output.argmax(dim=1).long() == batch_labels.long()). \\\n",
        "                float().sum().item()\n",
        "            iter_acc = float(iter_true_sample) / output.shape[0]\n",
        "            # print(\"Train accuracy: {acc}\".format(acc=iter_acc))\n",
        "            # utils.writer.add_scalar(\"iter train acc\", iter_acc, train_iter)\n",
        "            # print(f\"{train_iter} iter train acc: {iter_acc}\")\n",
        "            true_sample += iter_true_sample\n",
        "\n",
        "        acc = (true_sample / train_size)*100\n",
        "        # utils.writer.add_scalar(\"Epoch Acc\", acc, e)\n",
        "        # tqdm.write(f\"Epoch:{e}  \\t train_acc:{acc:.2f}\")\n",
        "        print(\"Train accuracy: {acc}\".format(acc=acc))\n",
        "#         plt.plot(ep_vt, epoch_losses_list, label='Train Loss')\n",
        "#         plt.legend()\n",
        "#         plt.show() \n",
        "\n",
        "        val_loss_list = []\n",
        "        # ep_vt = []\n",
        "        true_sample = 0\n",
        "        model.eval()\n",
        "        with torch.no_grad():\n",
        "            for i, (adj, features,batch_labels) in enumerate(val_loader):\n",
        "                # val_iter += 1\n",
        "                graph_feat, _ = model(features, adj)\n",
        "                output = model.classifier(graph_feat)\n",
        "                loss = model.loss(output, batch_labels)\n",
        "                val_loss_list.append(loss.item())\n",
        "                # ep_vt.append(val_iter)\n",
        "                iter_true_sample = (output.argmax(dim=1).long() == batch_labels.long()). \\\n",
        "                    float().sum().item()\n",
        "                iter_acc = float(iter_true_sample) / output.shape[0]\n",
        "                true_sample += iter_true_sample\n",
        "        acc = (true_sample / val_size)*100\n",
        "        mx_acc[0] = max(mx_acc[0],acc)\n",
        "#         print('Val Acc with Perturb: 0 :{}'.format(acc))\n",
        "\n",
        "#         if args.save_model is True:\n",
        "#           if (mx_acc >= 0.56) and (acc > mx_acc):\n",
        "#             torch.save(model.state_dict(), PATH + '/' + args.dataSet + \"/model_{}.torch\".format(version))\n",
        "#             version += 1\n",
        "\n",
        "        test_loss_list = []\n",
        "        true_sample = 0\n",
        "        model.eval()\n",
        "        with torch.no_grad():\n",
        "            for i, (adj, features,batch_labels) in enumerate(test_loader):\n",
        "                test_iter += 1\n",
        "                graph_feat, _ = model(features, adj)\n",
        "                output = model.classifier(graph_feat)\n",
        "                loss = model.loss(output, batch_labels)\n",
        "                test_loss_list.append(loss.item())\n",
        "                iter_true_sample = (output.argmax(dim=1).long() == batch_labels.long()). \\\n",
        "                    float().sum().item()\n",
        "                iter_acc = float(iter_true_sample) / output.shape[0]\n",
        "                # utils.writer.add_scalar(\"iter test acc\", iter_acc, test_iter)\n",
        "                # print(f\"{test_iter} iter test acc: {iter_acc}\")\n",
        "                true_sample += iter_true_sample\n",
        "        acc = (true_sample / test_size)*100\n",
        "        # mx_acc = max(mx_acc,acc)\n",
        "        en = time.time()\n",
        "#         print(\"Required Time: \",en-st)\n",
        "        # utils.writer.add_scalar(\"Epoch Acc\", acc, e)\n",
        "        # tqdm.write(f\"Epoch:{e}  \\t val_acc:{acc:.2f}\")\n",
        "#         print(\"Test accuracy with Perturb 0: {accuracy}\".format(accuracy=acc))\n",
        "        mx_acc[0] = max(mx_acc[0],acc)\n",
        "    \n",
        "#         mx_acc = 0\n",
        "        model.eval()\n",
        "        true_sample = 0\n",
        "        with torch.no_grad():\n",
        "            for i, (adj, features,batch_labels) in enumerate(val_loader):\n",
        "                # val_iter += 1\n",
        "                c_features = feature_perturb(features, perturb_ratio=0.5)\n",
        "                graph_feat, _ = model(c_features, adj)\n",
        "                output = model.classifier(graph_feat)\n",
        "                loss = model.loss(output, batch_labels)\n",
        "                val_loss_list.append(loss.item())\n",
        "                # ep_vt.append(val_iter)\n",
        "                iter_true_sample = (output.argmax(dim=1).long() == batch_labels.long()). \\\n",
        "                    float().sum().item()\n",
        "                iter_acc = float(iter_true_sample) / output.shape[0]\n",
        "                true_sample += iter_true_sample\n",
        "        acc = (true_sample / val_size)*100\n",
        "        mx_acc[1] = max(mx_acc[1],acc)\n",
        "#         print('Val Acc with  {}'.format(acc))\n",
        "\n",
        "        # if args.save_model is True:\n",
        "        #   if (mx_acc >= 0.65) and (acc > mx_acc):\n",
        "        #     torch.save(model.state_dict(), PATH + '/' + args.dataSet + \"/model_{}.torch\".format(version))\n",
        "        #     version += 1\n",
        "\n",
        "        test_loss_list = []\n",
        "        true_sample = 0\n",
        "        model.eval()\n",
        "        with torch.no_grad():\n",
        "            for i, (adj, features,batch_labels) in enumerate(test_loader):\n",
        "                test_iter += 1\n",
        "                c_features = feature_perturb(features, perturb_ratio=0.5)\n",
        "                graph_feat, _ = model(features, adj)\n",
        "                output = model.classifier(graph_feat)\n",
        "                loss = model.loss(output, batch_labels)\n",
        "                test_loss_list.append(loss.item())\n",
        "                iter_true_sample = (output.argmax(dim=1).long() == batch_labels.long()). \\\n",
        "                    float().sum().item()\n",
        "                iter_acc = float(iter_true_sample) / output.shape[0]\n",
        "                # utils.writer.add_scalar(\"iter test acc\", iter_acc, test_iter)\n",
        "                # print(f\"{test_iter} iter test acc: {iter_acc}\")\n",
        "                true_sample += iter_true_sample\n",
        "        acc = (true_sample / test_size)*100\n",
        "        mx_acc[1] = max(mx_acc[1],acc)\n",
        "        en = time.time()\n",
        "#         print(\"Required Time: \",en-st)\n",
        "        # utils.writer.add_scalar(\"Epoch Acc\", acc, e)\n",
        "        # tqdm.write(f\"Epoch:{e}  \\t val_acc:{acc:.2f}\")\n",
        "#         print(\"Test accuracy: {accuracy}\".format(accuracy=acc))\n",
        "\n",
        "\n",
        "#         mx_acc = 0\n",
        "        model.eval()\n",
        "        true_sample = 0\n",
        "        with torch.no_grad():\n",
        "            for i, (adj, features,batch_labels) in enumerate(val_loader):\n",
        "                # val_iter += 1\n",
        "                c_features = feature_perturb(features, perturb_ratio=1.0)\n",
        "                graph_feat, _ = model(c_features, adj)\n",
        "                output = model.classifier(graph_feat)\n",
        "                loss = model.loss(output, batch_labels)\n",
        "                val_loss_list.append(loss.item())\n",
        "                # ep_vt.append(val_iter)\n",
        "                iter_true_sample = (output.argmax(dim=1).long() == batch_labels.long()). \\\n",
        "                    float().sum().item()\n",
        "                iter_acc = float(iter_true_sample) / output.shape[0]\n",
        "                true_sample += iter_true_sample\n",
        "        acc = (true_sample / val_size)*100\n",
        "        mx_acc[2] = max(mx_acc[2],acc)\n",
        "#         print('Val Acc {}'.format(acc))\n",
        "\n",
        "        # if args.save_model is True:\n",
        "        #   if (mx_acc >= 0.65) and (acc > mx_acc):\n",
        "        #     torch.save(model.state_dict(), PATH + '/' + args.dataSet + \"/model_{}.torch\".format(version))\n",
        "        #     version += 1\n",
        "\n",
        "        test_loss_list = []\n",
        "        true_sample = 0\n",
        "        model.eval()\n",
        "        true_sample = 0\n",
        "        with torch.no_grad():\n",
        "            for i, (adj, features,batch_labels) in enumerate(test_loader):\n",
        "                test_iter += 1\n",
        "                c_features = feature_perturb(features, perturb_ratio=1.0)\n",
        "                graph_feat, _ = model(features, adj)\n",
        "                output = model.classifier(graph_feat)\n",
        "                loss = model.loss(output, batch_labels)\n",
        "                test_loss_list.append(loss.item())\n",
        "                iter_true_sample = (output.argmax(dim=1).long() == batch_labels.long()). \\\n",
        "                    float().sum().item()\n",
        "                iter_acc = float(iter_true_sample) / output.shape[0]\n",
        "                # utils.writer.add_scalar(\"iter test acc\", iter_acc, test_iter)\n",
        "                # print(f\"{test_iter} iter test acc: {iter_acc}\")\n",
        "                true_sample += iter_true_sample\n",
        "        acc = (true_sample / test_size)*100\n",
        "        mx_acc[2] = max(mx_acc[2],acc)\n",
        "        en = time.time()\n",
        "#         print(\"Required Time: \",en-st)\n",
        "        # utils.writer.add_scalar(\"Epoch Acc\", acc, e)\n",
        "        # tqdm.write(f\"Epoch:{e}  \\t val_acc:{acc:.2f}\")\n",
        "#         print(\"Test accuracy: {accuracy}\".format(accuracy=acc))\n",
        "\n",
        "#         mx_acc = 0\n",
        "        model.eval()\n",
        "        true_sample = 0\n",
        "        with torch.no_grad():\n",
        "            for i, (adj, features,batch_labels) in enumerate(val_loader):\n",
        "                # val_iter += 1\n",
        "                c_features = feature_perturb(features, perturb_ratio=1.5)\n",
        "                graph_feat, _ = model(c_features, adj)\n",
        "                output = model.classifier(graph_feat)\n",
        "                loss = model.loss(output, batch_labels)\n",
        "                val_loss_list.append(loss.item())\n",
        "                # ep_vt.append(val_iter)\n",
        "                iter_true_sample = (output.argmax(dim=1).long() == batch_labels.long()). \\\n",
        "                    float().sum().item()\n",
        "                iter_acc = float(iter_true_sample) / output.shape[0]\n",
        "                true_sample += iter_true_sample\n",
        "        acc = (true_sample / val_size)*100\n",
        "        mx_acc[3] = max(mx_acc[3],acc)\n",
        "#         print('Val Acc {}'.format(acc))\n",
        "\n",
        "        # if args.save_model is True:\n",
        "        #   if (mx_acc >= 0.65) and (acc > mx_acc):\n",
        "        #     torch.save(model.state_dict(), PATH + '/' + args.dataSet + \"/model_{}.torch\".format(version))\n",
        "        #     version += 1\n",
        "\n",
        "        test_loss_list = []\n",
        "        true_sample = 0\n",
        "        model.eval()\n",
        "        true_sample = 0\n",
        "        with torch.no_grad():\n",
        "            for i, (adj, features,batch_labels) in enumerate(test_loader):\n",
        "                test_iter += 1\n",
        "                c_features = feature_perturb(features, perturb_ratio=1.5)\n",
        "                graph_feat, _ = model(features, adj)\n",
        "                output = model.classifier(graph_feat)\n",
        "                loss = model.loss(output, batch_labels)\n",
        "                test_loss_list.append(loss.item())\n",
        "                iter_true_sample = (output.argmax(dim=1).long() == batch_labels.long()). \\\n",
        "                    float().sum().item()\n",
        "                iter_acc = float(iter_true_sample) / output.shape[0]\n",
        "                # utils.writer.add_scalar(\"iter test acc\", iter_acc, test_iter)\n",
        "                # print(f\"{test_iter} iter test acc: {iter_acc}\")\n",
        "                true_sample += iter_true_sample\n",
        "        acc = (true_sample / test_size)*100\n",
        "        mx_acc[3] = max(mx_acc[3],acc)\n",
        "        en = time.time()\n",
        "#         print(\"Required Time: \",en-st)\n",
        "        # utils.writer.add_scalar(\"Epoch Acc\", acc, e)\n",
        "        # tqdm.write(f\"Epoch:{e}  \\t val_acc:{acc:.2f}\")\n",
        "#         print(\"Test accuracy: {accuracy}\".format(accuracy=acc))\n",
        "\n",
        "#         mx_acc = 0\n",
        "        model.eval()\n",
        "        true_sample = 0\n",
        "        with torch.no_grad():\n",
        "            for i, (adj, features,batch_labels) in enumerate(val_loader):\n",
        "                # val_iter += 1\n",
        "                c_features = feature_perturb(features, perturb_ratio=2.0)\n",
        "                graph_feat, _ = model(c_features, adj)\n",
        "                output = model.classifier(graph_feat)\n",
        "                loss = model.loss(output, batch_labels)\n",
        "                val_loss_list.append(loss.item())\n",
        "                # ep_vt.append(val_iter)\n",
        "                iter_true_sample = (output.argmax(dim=1).long() == batch_labels.long()). \\\n",
        "                    float().sum().item()\n",
        "                iter_acc = float(iter_true_sample) / output.shape[0]\n",
        "                true_sample += iter_true_sample\n",
        "        acc = (true_sample / val_size)*100\n",
        "        mx_acc[4] = max(mx_acc[4],acc)\n",
        "#         print('Val Acc {}'.format(acc))\n",
        "\n",
        "        # if args.save_model is True:\n",
        "        #   if (mx_acc >= 0.65) and (acc > mx_acc):\n",
        "        #     torch.save(model.state_dict(), PATH + '/' + args.dataSet + \"/model_{}.torch\".format(version))\n",
        "        #     version += 1\n",
        "\n",
        "        test_loss_list = []\n",
        "        true_sample = 0\n",
        "        model.eval()\n",
        "        with torch.no_grad():\n",
        "            for i, (adj, features,batch_labels) in enumerate(test_loader):\n",
        "                test_iter += 1\n",
        "                c_features = feature_perturb(features, perturb_ratio=2.0)\n",
        "                graph_feat, _ = model(features, adj)\n",
        "                output = model.classifier(graph_feat)\n",
        "                loss = model.loss(output, batch_labels)\n",
        "                test_loss_list.append(loss.item())\n",
        "                iter_true_sample = (output.argmax(dim=1).long() == batch_labels.long()). \\\n",
        "                    float().sum().item()\n",
        "                iter_acc = float(iter_true_sample) / output.shape[0]\n",
        "                # utils.writer.add_scalar(\"iter test acc\", iter_acc, test_iter)\n",
        "                # print(f\"{test_iter} iter test acc: {iter_acc}\")\n",
        "                true_sample += iter_true_sample\n",
        "        acc = (true_sample / test_size)*100\n",
        "        mx_acc[4] = max(mx_acc[4],acc)\n",
        "        en = time.time()\n",
        "#         print(\"Required Time: \",en-st)\n",
        "        # utils.writer.add_scalar(\"Epoch Acc\", acc, e)\n",
        "        # tqdm.write(f\"Epoch:{e}  \\t val_acc:{acc:.2f}\")\n",
        "#         print(\"Test accuracy: {accuracy}\".format(accuracy=acc))\n",
        "        results = []\n",
        "        results.append(round(mx_acc[0], 4))\n",
        "        results.append(round(mx_acc[1], 4))\n",
        "        results.append(round(mx_acc[2], 4))\n",
        "        results.append(round(mx_acc[3], 4))\n",
        "        results.append(round(mx_acc[4], 4))\n",
        "        \n",
        "        print(\"Best accuracy without perturbation: \",results[0])\n",
        "        print(\"Best accuracy with 0.5 perturbation: \",results[1])\n",
        "        print(\"Best accuracy with 1.0 perturbation: \",results[2])\n",
        "        print(\"Best accuracy with 1.5 perturbation: \",results[3])\n",
        "        print(\"Best accuracy with 2.0 perturbation: \",results[4])\n",
        "    result_file = ffopen(file_name, 'a',title)\n",
        "    buffer = [str(d) for d in results]\n",
        "    buf = ','.join(buffer) + '\\n'\n",
        "    result_file.write(buf)\n",
        "    result_file.close()\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Kio9wY1OYPqF"
      },
      "source": [
        "# File Management"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DGHuyURNYO6U"
      },
      "source": [
        "import time,os,errno\n",
        "            \n",
        "\n",
        "def ffopen(fileLocation, mode, title=None):\n",
        "  # if not os.path.exists(os.path.dirname(fileLocation)):\n",
        "  if not os.path.exists(fileLocation):\n",
        "    try:\n",
        "      os.makedirs(os.path.dirname(fileLocation))\n",
        "    except OSError as exc:  # Guard against race condition\n",
        "      if exc.errno != errno.EEXIST:\n",
        "        raise\n",
        "    f = open(fileLocation, mode)\n",
        "    if title is not None:\n",
        "      f.write(title)\n",
        "    f.close()\n",
        "\n",
        "  f = open(fileLocation, mode)\n",
        "  return f"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TNkSnm7J8SFs"
      },
      "source": [
        "#Execution"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "p5wb58mrsJda",
        "outputId": "a1da8cad-6159-4bd9-d377-4f6755536731"
      },
      "source": [
        "if __name__ == '__main__':\n",
        "  seed = 42\n",
        "  \n",
        "  random.seed(seed)\n",
        "\t\n",
        "  \n",
        "  np.random.seed(seed)\n",
        "\t\n",
        "  \n",
        "  torch.manual_seed(seed)\n",
        "\t\n",
        "  \n",
        "  torch.cuda.manual_seed_all(seed)\n",
        "  \n",
        "  \n",
        "  main()\n",
        "\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/dgl/base.py:25: UserWarning: Currently adjacency_matrix() returns a matrix with destination as rows by default.  In 0.5 the result will have source as rows (i.e. transpose=True)\n",
            "  warnings.warn(msg, warn_type)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch:  0\n",
            "Train accuracy: 19.375\n",
            "Best accuracy without perturbation:  25.0\n",
            "Best accuracy with 0.5 perturbation:  26.6667\n",
            "Best accuracy with 1.0 perturbation:  25.0\n",
            "Best accuracy with 1.5 perturbation:  26.6667\n",
            "Best accuracy with 2.0 perturbation:  25.0\n",
            "Epoch:  1\n",
            "Train accuracy: 18.75\n",
            "Best accuracy without perturbation:  25.0\n",
            "Best accuracy with 0.5 perturbation:  26.6667\n",
            "Best accuracy with 1.0 perturbation:  25.0\n",
            "Best accuracy with 1.5 perturbation:  26.6667\n",
            "Best accuracy with 2.0 perturbation:  25.0\n",
            "Epoch:  2\n",
            "Train accuracy: 21.875\n",
            "Best accuracy without perturbation:  28.3333\n",
            "Best accuracy with 0.5 perturbation:  26.6667\n",
            "Best accuracy with 1.0 perturbation:  25.0\n",
            "Best accuracy with 1.5 perturbation:  26.6667\n",
            "Best accuracy with 2.0 perturbation:  25.0\n",
            "Epoch:  3\n",
            "Train accuracy: 23.125\n",
            "Best accuracy without perturbation:  28.3333\n",
            "Best accuracy with 0.5 perturbation:  28.3333\n",
            "Best accuracy with 1.0 perturbation:  25.0\n",
            "Best accuracy with 1.5 perturbation:  26.6667\n",
            "Best accuracy with 2.0 perturbation:  25.0\n",
            "Epoch:  4\n",
            "Train accuracy: 22.083333333333332\n",
            "Best accuracy without perturbation:  28.3333\n",
            "Best accuracy with 0.5 perturbation:  28.3333\n",
            "Best accuracy with 1.0 perturbation:  25.0\n",
            "Best accuracy with 1.5 perturbation:  26.6667\n",
            "Best accuracy with 2.0 perturbation:  25.0\n",
            "Epoch:  5\n",
            "Train accuracy: 26.458333333333332\n",
            "Best accuracy without perturbation:  28.3333\n",
            "Best accuracy with 0.5 perturbation:  28.3333\n",
            "Best accuracy with 1.0 perturbation:  25.0\n",
            "Best accuracy with 1.5 perturbation:  26.6667\n",
            "Best accuracy with 2.0 perturbation:  28.3333\n",
            "Epoch:  6\n",
            "Train accuracy: 25.624999999999996\n",
            "Best accuracy without perturbation:  30.0\n",
            "Best accuracy with 0.5 perturbation:  28.3333\n",
            "Best accuracy with 1.0 perturbation:  25.0\n",
            "Best accuracy with 1.5 perturbation:  26.6667\n",
            "Best accuracy with 2.0 perturbation:  28.3333\n",
            "Epoch:  7\n",
            "Train accuracy: 28.749999999999996\n",
            "Best accuracy without perturbation:  31.6667\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  25.0\n",
            "Best accuracy with 1.5 perturbation:  26.6667\n",
            "Best accuracy with 2.0 perturbation:  28.3333\n",
            "Epoch:  8\n",
            "Train accuracy: 28.541666666666664\n",
            "Best accuracy without perturbation:  31.6667\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  25.0\n",
            "Best accuracy with 1.5 perturbation:  26.6667\n",
            "Best accuracy with 2.0 perturbation:  28.3333\n",
            "Epoch:  9\n",
            "Train accuracy: 28.125\n",
            "Best accuracy without perturbation:  31.6667\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  30.0\n",
            "Best accuracy with 1.5 perturbation:  26.6667\n",
            "Best accuracy with 2.0 perturbation:  28.3333\n",
            "Epoch:  10\n",
            "Train accuracy: 27.083333333333332\n",
            "Best accuracy without perturbation:  31.6667\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  30.0\n",
            "Best accuracy with 1.5 perturbation:  28.3333\n",
            "Best accuracy with 2.0 perturbation:  28.3333\n",
            "Epoch:  11\n",
            "Train accuracy: 31.458333333333332\n",
            "Best accuracy without perturbation:  31.6667\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  30.0\n",
            "Best accuracy with 1.5 perturbation:  28.3333\n",
            "Best accuracy with 2.0 perturbation:  28.3333\n",
            "Epoch:  12\n",
            "Train accuracy: 29.791666666666668\n",
            "Best accuracy without perturbation:  31.6667\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  30.0\n",
            "Best accuracy with 1.5 perturbation:  28.3333\n",
            "Best accuracy with 2.0 perturbation:  28.3333\n",
            "Epoch:  13\n",
            "Train accuracy: 33.125\n",
            "Best accuracy without perturbation:  31.6667\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  30.0\n",
            "Best accuracy with 1.5 perturbation:  28.3333\n",
            "Best accuracy with 2.0 perturbation:  28.3333\n",
            "Epoch:  14\n",
            "Train accuracy: 32.29166666666667\n",
            "Best accuracy without perturbation:  31.6667\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  30.0\n",
            "Best accuracy with 1.5 perturbation:  28.3333\n",
            "Best accuracy with 2.0 perturbation:  28.3333\n",
            "Epoch:  15\n",
            "Train accuracy: 32.916666666666664\n",
            "Best accuracy without perturbation:  31.6667\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  30.0\n",
            "Best accuracy with 1.5 perturbation:  28.3333\n",
            "Best accuracy with 2.0 perturbation:  28.3333\n",
            "Epoch:  16\n",
            "Train accuracy: 35.0\n",
            "Best accuracy without perturbation:  33.3333\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  30.0\n",
            "Best accuracy with 1.5 perturbation:  28.3333\n",
            "Best accuracy with 2.0 perturbation:  28.3333\n",
            "Epoch:  17\n",
            "Train accuracy: 35.625\n",
            "Best accuracy without perturbation:  33.3333\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  30.0\n",
            "Best accuracy with 1.5 perturbation:  28.3333\n",
            "Best accuracy with 2.0 perturbation:  28.3333\n",
            "Epoch:  18\n",
            "Train accuracy: 35.0\n",
            "Best accuracy without perturbation:  36.6667\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  30.0\n",
            "Best accuracy with 1.5 perturbation:  28.3333\n",
            "Best accuracy with 2.0 perturbation:  33.3333\n",
            "Epoch:  19\n",
            "Train accuracy: 36.25\n",
            "Best accuracy without perturbation:  36.6667\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  30.0\n",
            "Best accuracy with 1.5 perturbation:  28.3333\n",
            "Best accuracy with 2.0 perturbation:  33.3333\n",
            "Epoch:  20\n",
            "Train accuracy: 36.875\n",
            "Best accuracy without perturbation:  36.6667\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  31.6667\n",
            "Best accuracy with 1.5 perturbation:  31.6667\n",
            "Best accuracy with 2.0 perturbation:  33.3333\n",
            "Epoch:  21\n",
            "Train accuracy: 37.916666666666664\n",
            "Best accuracy without perturbation:  36.6667\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  31.6667\n",
            "Best accuracy with 1.5 perturbation:  31.6667\n",
            "Best accuracy with 2.0 perturbation:  33.3333\n",
            "Epoch:  22\n",
            "Train accuracy: 36.666666666666664\n",
            "Best accuracy without perturbation:  36.6667\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  31.6667\n",
            "Best accuracy with 1.5 perturbation:  31.6667\n",
            "Best accuracy with 2.0 perturbation:  33.3333\n",
            "Epoch:  23\n",
            "Train accuracy: 41.875\n",
            "Best accuracy without perturbation:  38.3333\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  31.6667\n",
            "Best accuracy with 1.5 perturbation:  31.6667\n",
            "Best accuracy with 2.0 perturbation:  33.3333\n",
            "Epoch:  24\n",
            "Train accuracy: 41.875\n",
            "Best accuracy without perturbation:  38.3333\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  31.6667\n",
            "Best accuracy with 1.5 perturbation:  33.3333\n",
            "Best accuracy with 2.0 perturbation:  33.3333\n",
            "Epoch:  25\n",
            "Train accuracy: 45.0\n",
            "Best accuracy without perturbation:  38.3333\n",
            "Best accuracy with 0.5 perturbation:  33.3333\n",
            "Best accuracy with 1.0 perturbation:  31.6667\n",
            "Best accuracy with 1.5 perturbation:  33.3333\n",
            "Best accuracy with 2.0 perturbation:  35.0\n",
            "Epoch:  26\n",
            "Train accuracy: 42.70833333333333\n",
            "Best accuracy without perturbation:  38.3333\n",
            "Best accuracy with 0.5 perturbation:  36.6667\n",
            "Best accuracy with 1.0 perturbation:  31.6667\n",
            "Best accuracy with 1.5 perturbation:  33.3333\n",
            "Best accuracy with 2.0 perturbation:  35.0\n",
            "Epoch:  27\n",
            "Train accuracy: 44.375\n",
            "Best accuracy without perturbation:  46.6667\n",
            "Best accuracy with 0.5 perturbation:  36.6667\n",
            "Best accuracy with 1.0 perturbation:  31.6667\n",
            "Best accuracy with 1.5 perturbation:  33.3333\n",
            "Best accuracy with 2.0 perturbation:  35.0\n",
            "Epoch:  28\n",
            "Train accuracy: 44.375\n",
            "Best accuracy without perturbation:  46.6667\n",
            "Best accuracy with 0.5 perturbation:  36.6667\n",
            "Best accuracy with 1.0 perturbation:  35.0\n",
            "Best accuracy with 1.5 perturbation:  35.0\n",
            "Best accuracy with 2.0 perturbation:  35.0\n",
            "Epoch:  29\n",
            "Train accuracy: 43.333333333333336\n",
            "Best accuracy without perturbation:  46.6667\n",
            "Best accuracy with 0.5 perturbation:  36.6667\n",
            "Best accuracy with 1.0 perturbation:  40.0\n",
            "Best accuracy with 1.5 perturbation:  36.6667\n",
            "Best accuracy with 2.0 perturbation:  40.0\n",
            "Epoch:  30\n",
            "Train accuracy: 49.375\n",
            "Best accuracy without perturbation:  50.0\n",
            "Best accuracy with 0.5 perturbation:  43.3333\n",
            "Best accuracy with 1.0 perturbation:  40.0\n",
            "Best accuracy with 1.5 perturbation:  36.6667\n",
            "Best accuracy with 2.0 perturbation:  40.0\n",
            "Epoch:  31\n",
            "Train accuracy: 47.5\n",
            "Best accuracy without perturbation:  50.0\n",
            "Best accuracy with 0.5 perturbation:  43.3333\n",
            "Best accuracy with 1.0 perturbation:  40.0\n",
            "Best accuracy with 1.5 perturbation:  36.6667\n",
            "Best accuracy with 2.0 perturbation:  40.0\n",
            "Epoch:  32\n",
            "Train accuracy: 47.083333333333336\n",
            "Best accuracy without perturbation:  50.0\n",
            "Best accuracy with 0.5 perturbation:  43.3333\n",
            "Best accuracy with 1.0 perturbation:  40.0\n",
            "Best accuracy with 1.5 perturbation:  36.6667\n",
            "Best accuracy with 2.0 perturbation:  40.0\n",
            "Epoch:  33\n",
            "Train accuracy: 47.91666666666667\n",
            "Best accuracy without perturbation:  50.0\n",
            "Best accuracy with 0.5 perturbation:  43.3333\n",
            "Best accuracy with 1.0 perturbation:  40.0\n",
            "Best accuracy with 1.5 perturbation:  36.6667\n",
            "Best accuracy with 2.0 perturbation:  40.0\n",
            "Epoch:  34\n",
            "Train accuracy: 53.541666666666664\n",
            "Best accuracy without perturbation:  50.0\n",
            "Best accuracy with 0.5 perturbation:  43.3333\n",
            "Best accuracy with 1.0 perturbation:  40.0\n",
            "Best accuracy with 1.5 perturbation:  36.6667\n",
            "Best accuracy with 2.0 perturbation:  43.3333\n",
            "Epoch:  35\n",
            "Train accuracy: 53.541666666666664\n",
            "Best accuracy without perturbation:  51.6667\n",
            "Best accuracy with 0.5 perturbation:  43.3333\n",
            "Best accuracy with 1.0 perturbation:  40.0\n",
            "Best accuracy with 1.5 perturbation:  36.6667\n",
            "Best accuracy with 2.0 perturbation:  43.3333\n",
            "Epoch:  36\n",
            "Train accuracy: 52.916666666666664\n",
            "Best accuracy without perturbation:  51.6667\n",
            "Best accuracy with 0.5 perturbation:  43.3333\n",
            "Best accuracy with 1.0 perturbation:  40.0\n",
            "Best accuracy with 1.5 perturbation:  40.0\n",
            "Best accuracy with 2.0 perturbation:  43.3333\n",
            "Epoch:  37\n",
            "Train accuracy: 52.5\n",
            "Best accuracy without perturbation:  51.6667\n",
            "Best accuracy with 0.5 perturbation:  43.3333\n",
            "Best accuracy with 1.0 perturbation:  43.3333\n",
            "Best accuracy with 1.5 perturbation:  40.0\n",
            "Best accuracy with 2.0 perturbation:  43.3333\n",
            "Epoch:  38\n",
            "Train accuracy: 52.916666666666664\n",
            "Best accuracy without perturbation:  51.6667\n",
            "Best accuracy with 0.5 perturbation:  43.3333\n",
            "Best accuracy with 1.0 perturbation:  43.3333\n",
            "Best accuracy with 1.5 perturbation:  40.0\n",
            "Best accuracy with 2.0 perturbation:  43.3333\n",
            "Epoch:  39\n",
            "Train accuracy: 52.29166666666667\n",
            "Best accuracy without perturbation:  51.6667\n",
            "Best accuracy with 0.5 perturbation:  43.3333\n",
            "Best accuracy with 1.0 perturbation:  43.3333\n",
            "Best accuracy with 1.5 perturbation:  41.6667\n",
            "Best accuracy with 2.0 perturbation:  43.3333\n",
            "Epoch:  40\n",
            "Train accuracy: 56.666666666666664\n",
            "Best accuracy without perturbation:  51.6667\n",
            "Best accuracy with 0.5 perturbation:  43.3333\n",
            "Best accuracy with 1.0 perturbation:  43.3333\n",
            "Best accuracy with 1.5 perturbation:  41.6667\n",
            "Best accuracy with 2.0 perturbation:  45.0\n",
            "Epoch:  41\n",
            "Train accuracy: 60.83333333333333\n",
            "Best accuracy without perturbation:  51.6667\n",
            "Best accuracy with 0.5 perturbation:  43.3333\n",
            "Best accuracy with 1.0 perturbation:  43.3333\n",
            "Best accuracy with 1.5 perturbation:  46.6667\n",
            "Best accuracy with 2.0 perturbation:  45.0\n",
            "Epoch:  42\n",
            "Train accuracy: 61.458333333333336\n",
            "Best accuracy without perturbation:  51.6667\n",
            "Best accuracy with 0.5 perturbation:  43.3333\n",
            "Best accuracy with 1.0 perturbation:  45.0\n",
            "Best accuracy with 1.5 perturbation:  46.6667\n",
            "Best accuracy with 2.0 perturbation:  45.0\n",
            "Epoch:  43\n",
            "Train accuracy: 62.916666666666664\n",
            "Best accuracy without perturbation:  53.3333\n",
            "Best accuracy with 0.5 perturbation:  43.3333\n",
            "Best accuracy with 1.0 perturbation:  45.0\n",
            "Best accuracy with 1.5 perturbation:  46.6667\n",
            "Best accuracy with 2.0 perturbation:  46.6667\n",
            "Epoch:  44\n",
            "Train accuracy: 62.083333333333336\n",
            "Best accuracy without perturbation:  53.3333\n",
            "Best accuracy with 0.5 perturbation:  50.0\n",
            "Best accuracy with 1.0 perturbation:  45.0\n",
            "Best accuracy with 1.5 perturbation:  46.6667\n",
            "Best accuracy with 2.0 perturbation:  46.6667\n",
            "Epoch:  45\n",
            "Train accuracy: 64.16666666666667\n",
            "Best accuracy without perturbation:  53.3333\n",
            "Best accuracy with 0.5 perturbation:  50.0\n",
            "Best accuracy with 1.0 perturbation:  45.0\n",
            "Best accuracy with 1.5 perturbation:  46.6667\n",
            "Best accuracy with 2.0 perturbation:  46.6667\n",
            "Epoch:  46\n",
            "Train accuracy: 61.25000000000001\n",
            "Best accuracy without perturbation:  53.3333\n",
            "Best accuracy with 0.5 perturbation:  50.0\n",
            "Best accuracy with 1.0 perturbation:  45.0\n",
            "Best accuracy with 1.5 perturbation:  46.6667\n",
            "Best accuracy with 2.0 perturbation:  46.6667\n",
            "Epoch:  47\n",
            "Train accuracy: 64.58333333333334\n",
            "Best accuracy without perturbation:  53.3333\n",
            "Best accuracy with 0.5 perturbation:  50.0\n",
            "Best accuracy with 1.0 perturbation:  45.0\n",
            "Best accuracy with 1.5 perturbation:  46.6667\n",
            "Best accuracy with 2.0 perturbation:  46.6667\n",
            "Epoch:  48\n",
            "Train accuracy: 65.0\n",
            "Best accuracy without perturbation:  53.3333\n",
            "Best accuracy with 0.5 perturbation:  50.0\n",
            "Best accuracy with 1.0 perturbation:  45.0\n",
            "Best accuracy with 1.5 perturbation:  46.6667\n",
            "Best accuracy with 2.0 perturbation:  46.6667\n",
            "Epoch:  49\n",
            "Train accuracy: 66.25\n",
            "Best accuracy without perturbation:  55.0\n",
            "Best accuracy with 0.5 perturbation:  50.0\n",
            "Best accuracy with 1.0 perturbation:  45.0\n",
            "Best accuracy with 1.5 perturbation:  46.6667\n",
            "Best accuracy with 2.0 perturbation:  46.6667\n",
            "Epoch:  50\n",
            "Train accuracy: 65.20833333333333\n",
            "Best accuracy without perturbation:  55.0\n",
            "Best accuracy with 0.5 perturbation:  50.0\n",
            "Best accuracy with 1.0 perturbation:  45.0\n",
            "Best accuracy with 1.5 perturbation:  46.6667\n",
            "Best accuracy with 2.0 perturbation:  46.6667\n",
            "Epoch:  51\n",
            "Train accuracy: 68.33333333333333\n",
            "Best accuracy without perturbation:  55.0\n",
            "Best accuracy with 0.5 perturbation:  50.0\n",
            "Best accuracy with 1.0 perturbation:  45.0\n",
            "Best accuracy with 1.5 perturbation:  46.6667\n",
            "Best accuracy with 2.0 perturbation:  46.6667\n",
            "Epoch:  52\n",
            "Train accuracy: 66.66666666666666\n",
            "Best accuracy without perturbation:  55.0\n",
            "Best accuracy with 0.5 perturbation:  50.0\n",
            "Best accuracy with 1.0 perturbation:  46.6667\n",
            "Best accuracy with 1.5 perturbation:  46.6667\n",
            "Best accuracy with 2.0 perturbation:  46.6667\n",
            "Epoch:  53\n",
            "Train accuracy: 68.54166666666667\n",
            "Best accuracy without perturbation:  56.6667\n",
            "Best accuracy with 0.5 perturbation:  53.3333\n",
            "Best accuracy with 1.0 perturbation:  53.3333\n",
            "Best accuracy with 1.5 perturbation:  46.6667\n",
            "Best accuracy with 2.0 perturbation:  50.0\n",
            "Epoch:  54\n",
            "Train accuracy: 71.04166666666667\n",
            "Best accuracy without perturbation:  56.6667\n",
            "Best accuracy with 0.5 perturbation:  53.3333\n",
            "Best accuracy with 1.0 perturbation:  53.3333\n",
            "Best accuracy with 1.5 perturbation:  46.6667\n",
            "Best accuracy with 2.0 perturbation:  50.0\n",
            "Epoch:  55\n",
            "Train accuracy: 73.125\n",
            "Best accuracy without perturbation:  56.6667\n",
            "Best accuracy with 0.5 perturbation:  53.3333\n",
            "Best accuracy with 1.0 perturbation:  53.3333\n",
            "Best accuracy with 1.5 perturbation:  46.6667\n",
            "Best accuracy with 2.0 perturbation:  51.6667\n",
            "Epoch:  56\n",
            "Train accuracy: 70.83333333333334\n",
            "Best accuracy without perturbation:  56.6667\n",
            "Best accuracy with 0.5 perturbation:  53.3333\n",
            "Best accuracy with 1.0 perturbation:  53.3333\n",
            "Best accuracy with 1.5 perturbation:  50.0\n",
            "Best accuracy with 2.0 perturbation:  51.6667\n",
            "Epoch:  57\n",
            "Train accuracy: 73.75\n",
            "Best accuracy without perturbation:  56.6667\n",
            "Best accuracy with 0.5 perturbation:  53.3333\n",
            "Best accuracy with 1.0 perturbation:  53.3333\n",
            "Best accuracy with 1.5 perturbation:  50.0\n",
            "Best accuracy with 2.0 perturbation:  51.6667\n",
            "Epoch:  58\n",
            "Train accuracy: 74.79166666666667\n",
            "Best accuracy without perturbation:  56.6667\n",
            "Best accuracy with 0.5 perturbation:  53.3333\n",
            "Best accuracy with 1.0 perturbation:  53.3333\n",
            "Best accuracy with 1.5 perturbation:  50.0\n",
            "Best accuracy with 2.0 perturbation:  51.6667\n",
            "Epoch:  59\n",
            "Train accuracy: 76.25\n",
            "Best accuracy without perturbation:  56.6667\n",
            "Best accuracy with 0.5 perturbation:  53.3333\n",
            "Best accuracy with 1.0 perturbation:  53.3333\n",
            "Best accuracy with 1.5 perturbation:  50.0\n",
            "Best accuracy with 2.0 perturbation:  51.6667\n",
            "Epoch:  60\n",
            "Train accuracy: 75.0\n",
            "Best accuracy without perturbation:  56.6667\n",
            "Best accuracy with 0.5 perturbation:  53.3333\n",
            "Best accuracy with 1.0 perturbation:  53.3333\n",
            "Best accuracy with 1.5 perturbation:  50.0\n",
            "Best accuracy with 2.0 perturbation:  51.6667\n",
            "Epoch:  61\n",
            "Train accuracy: 75.41666666666667\n",
            "Best accuracy without perturbation:  56.6667\n",
            "Best accuracy with 0.5 perturbation:  53.3333\n",
            "Best accuracy with 1.0 perturbation:  53.3333\n",
            "Best accuracy with 1.5 perturbation:  50.0\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  62\n",
            "Train accuracy: 76.45833333333333\n",
            "Best accuracy without perturbation:  56.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  53.3333\n",
            "Best accuracy with 1.5 perturbation:  50.0\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  63\n",
            "Train accuracy: 76.25\n",
            "Best accuracy without perturbation:  56.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  53.3333\n",
            "Best accuracy with 1.5 perturbation:  50.0\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  64\n",
            "Train accuracy: 73.95833333333334\n",
            "Best accuracy without perturbation:  56.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  53.3333\n",
            "Best accuracy with 1.5 perturbation:  50.0\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  65\n",
            "Train accuracy: 73.54166666666667\n",
            "Best accuracy without perturbation:  56.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  53.3333\n",
            "Best accuracy with 1.5 perturbation:  50.0\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  66\n",
            "Train accuracy: 78.75\n",
            "Best accuracy without perturbation:  56.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  53.3333\n",
            "Best accuracy with 1.5 perturbation:  50.0\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  67\n",
            "Train accuracy: 75.0\n",
            "Best accuracy without perturbation:  56.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  53.3333\n",
            "Best accuracy with 1.5 perturbation:  51.6667\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  68\n",
            "Train accuracy: 76.25\n",
            "Best accuracy without perturbation:  58.3333\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  55.0\n",
            "Best accuracy with 1.5 perturbation:  51.6667\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  69\n",
            "Train accuracy: 76.45833333333333\n",
            "Best accuracy without perturbation:  58.3333\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  55.0\n",
            "Best accuracy with 1.5 perturbation:  51.6667\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  70\n",
            "Train accuracy: 78.75\n",
            "Best accuracy without perturbation:  58.3333\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  51.6667\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  71\n",
            "Train accuracy: 76.25\n",
            "Best accuracy without perturbation:  58.3333\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  51.6667\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  72\n",
            "Train accuracy: 77.29166666666667\n",
            "Best accuracy without perturbation:  58.3333\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  51.6667\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  73\n",
            "Train accuracy: 81.25\n",
            "Best accuracy without perturbation:  58.3333\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  56.6667\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  74\n",
            "Train accuracy: 81.66666666666667\n",
            "Best accuracy without perturbation:  58.3333\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  56.6667\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  75\n",
            "Train accuracy: 81.45833333333333\n",
            "Best accuracy without perturbation:  63.3333\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  60.0\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  76\n",
            "Train accuracy: 77.29166666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  60.0\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  77\n",
            "Train accuracy: 85.20833333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  60.0\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  78\n",
            "Train accuracy: 80.625\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  60.0\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  79\n",
            "Train accuracy: 82.08333333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  60.0\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  80\n",
            "Train accuracy: 81.66666666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  60.0\n",
            "Best accuracy with 2.0 perturbation:  56.6667\n",
            "Epoch:  81\n",
            "Train accuracy: 82.70833333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  60.0\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  82\n",
            "Train accuracy: 84.79166666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  60.0\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  83\n",
            "Train accuracy: 83.54166666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  60.0\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  84\n",
            "Train accuracy: 84.375\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  60.0\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  85\n",
            "Train accuracy: 82.70833333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  60.0\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  86\n",
            "Train accuracy: 81.04166666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  87\n",
            "Train accuracy: 86.04166666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  88\n",
            "Train accuracy: 83.33333333333334\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  89\n",
            "Train accuracy: 85.20833333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  90\n",
            "Train accuracy: 86.25\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  91\n",
            "Train accuracy: 86.45833333333334\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  92\n",
            "Train accuracy: 82.70833333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  93\n",
            "Train accuracy: 87.08333333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  94\n",
            "Train accuracy: 85.83333333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  95\n",
            "Train accuracy: 84.79166666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  96\n",
            "Train accuracy: 88.75\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  58.3333\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  97\n",
            "Train accuracy: 86.45833333333334\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  60.0\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  98\n",
            "Train accuracy: 86.04166666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  60.0\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  99\n",
            "Train accuracy: 88.75\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  60.0\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  100\n",
            "Train accuracy: 90.20833333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  60.0\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  101\n",
            "Train accuracy: 87.91666666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  60.0\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  102\n",
            "Train accuracy: 87.5\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  60.0\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  103\n",
            "Train accuracy: 88.95833333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  60.0\n",
            "Best accuracy with 1.5 perturbation:  61.6667\n",
            "Best accuracy with 2.0 perturbation:  60.0\n",
            "Epoch:  104\n",
            "Train accuracy: 88.125\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  60.0\n",
            "Best accuracy with 1.5 perturbation:  63.3333\n",
            "Best accuracy with 2.0 perturbation:  61.6667\n",
            "Epoch:  105\n",
            "Train accuracy: 90.625\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  60.0\n",
            "Best accuracy with 1.5 perturbation:  63.3333\n",
            "Best accuracy with 2.0 perturbation:  61.6667\n",
            "Epoch:  106\n",
            "Train accuracy: 88.54166666666666\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  60.0\n",
            "Best accuracy with 1.5 perturbation:  63.3333\n",
            "Best accuracy with 2.0 perturbation:  61.6667\n",
            "Epoch:  107\n",
            "Train accuracy: 88.95833333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  58.3333\n",
            "Best accuracy with 1.0 perturbation:  60.0\n",
            "Best accuracy with 1.5 perturbation:  63.3333\n",
            "Best accuracy with 2.0 perturbation:  61.6667\n",
            "Epoch:  108\n",
            "Train accuracy: 88.95833333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  60.0\n",
            "Best accuracy with 1.0 perturbation:  60.0\n",
            "Best accuracy with 1.5 perturbation:  63.3333\n",
            "Best accuracy with 2.0 perturbation:  61.6667\n",
            "Epoch:  109\n",
            "Train accuracy: 91.25\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  60.0\n",
            "Best accuracy with 1.0 perturbation:  61.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  61.6667\n",
            "Epoch:  110\n",
            "Train accuracy: 88.75\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  60.0\n",
            "Best accuracy with 1.0 perturbation:  61.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  61.6667\n",
            "Epoch:  111\n",
            "Train accuracy: 91.04166666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  60.0\n",
            "Best accuracy with 1.0 perturbation:  61.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  63.3333\n",
            "Epoch:  112\n",
            "Train accuracy: 92.08333333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  60.0\n",
            "Best accuracy with 1.0 perturbation:  61.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  63.3333\n",
            "Epoch:  113\n",
            "Train accuracy: 89.375\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  60.0\n",
            "Best accuracy with 1.0 perturbation:  61.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  63.3333\n",
            "Epoch:  114\n",
            "Train accuracy: 89.79166666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  60.0\n",
            "Best accuracy with 1.0 perturbation:  61.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  63.3333\n",
            "Epoch:  115\n",
            "Train accuracy: 89.375\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  60.0\n",
            "Best accuracy with 1.0 perturbation:  61.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  116\n",
            "Train accuracy: 91.45833333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  60.0\n",
            "Best accuracy with 1.0 perturbation:  61.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  117\n",
            "Train accuracy: 91.04166666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  60.0\n",
            "Best accuracy with 1.0 perturbation:  61.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  118\n",
            "Train accuracy: 90.83333333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  60.0\n",
            "Best accuracy with 1.0 perturbation:  61.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  119\n",
            "Train accuracy: 90.41666666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  120\n",
            "Train accuracy: 89.375\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  121\n",
            "Train accuracy: 92.70833333333334\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  122\n",
            "Train accuracy: 93.95833333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  123\n",
            "Train accuracy: 92.91666666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  124\n",
            "Train accuracy: 94.16666666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  125\n",
            "Train accuracy: 92.91666666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  126\n",
            "Train accuracy: 92.29166666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  127\n",
            "Train accuracy: 92.29166666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  128\n",
            "Train accuracy: 89.79166666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  129\n",
            "Train accuracy: 93.54166666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  130\n",
            "Train accuracy: 92.08333333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  131\n",
            "Train accuracy: 93.33333333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  132\n",
            "Train accuracy: 91.875\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  133\n",
            "Train accuracy: 92.70833333333334\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  134\n",
            "Train accuracy: 93.125\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  135\n",
            "Train accuracy: 95.20833333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  136\n",
            "Train accuracy: 93.54166666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  137\n",
            "Train accuracy: 92.08333333333333\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  65.0\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  138\n",
            "Train accuracy: 94.79166666666666\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  139\n",
            "Train accuracy: 93.54166666666667\n",
            "Best accuracy without perturbation:  66.6667\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  140\n",
            "Train accuracy: 93.33333333333333\n",
            "Best accuracy without perturbation:  68.3333\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  141\n",
            "Train accuracy: 94.58333333333333\n",
            "Best accuracy without perturbation:  68.3333\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  142\n",
            "Train accuracy: 96.45833333333333\n",
            "Best accuracy without perturbation:  68.3333\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  143\n",
            "Train accuracy: 94.58333333333333\n",
            "Best accuracy without perturbation:  68.3333\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  144\n",
            "Train accuracy: 93.75\n",
            "Best accuracy without perturbation:  68.3333\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  145\n",
            "Train accuracy: 96.875\n",
            "Best accuracy without perturbation:  68.3333\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  146\n",
            "Train accuracy: 95.20833333333333\n",
            "Best accuracy without perturbation:  68.3333\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  147\n",
            "Train accuracy: 95.0\n",
            "Best accuracy without perturbation:  68.3333\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  148\n",
            "Train accuracy: 93.75\n",
            "Best accuracy without perturbation:  68.3333\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  149\n",
            "Train accuracy: 97.5\n",
            "Best accuracy without perturbation:  68.3333\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  150\n",
            "Train accuracy: 96.25\n",
            "Best accuracy without perturbation:  68.3333\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  151\n",
            "Train accuracy: 96.45833333333333\n",
            "Best accuracy without perturbation:  68.3333\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  152\n",
            "Train accuracy: 96.45833333333333\n",
            "Best accuracy without perturbation:  68.3333\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  153\n",
            "Train accuracy: 96.875\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  154\n",
            "Train accuracy: 94.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  155\n",
            "Train accuracy: 94.79166666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  156\n",
            "Train accuracy: 94.79166666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  157\n",
            "Train accuracy: 95.41666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  158\n",
            "Train accuracy: 96.25\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  159\n",
            "Train accuracy: 95.625\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  160\n",
            "Train accuracy: 91.66666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  161\n",
            "Train accuracy: 96.875\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  162\n",
            "Train accuracy: 95.83333333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  163\n",
            "Train accuracy: 96.04166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  164\n",
            "Train accuracy: 94.79166666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  165\n",
            "Train accuracy: 97.70833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  166\n",
            "Train accuracy: 95.625\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  167\n",
            "Train accuracy: 96.04166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  168\n",
            "Train accuracy: 96.66666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  169\n",
            "Train accuracy: 96.875\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  170\n",
            "Train accuracy: 95.41666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  171\n",
            "Train accuracy: 96.45833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  172\n",
            "Train accuracy: 97.08333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  173\n",
            "Train accuracy: 97.5\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  174\n",
            "Train accuracy: 97.29166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  175\n",
            "Train accuracy: 97.08333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  176\n",
            "Train accuracy: 95.0\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  177\n",
            "Train accuracy: 97.5\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  178\n",
            "Train accuracy: 97.29166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  65.0\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  179\n",
            "Train accuracy: 95.41666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  180\n",
            "Train accuracy: 95.83333333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  181\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  182\n",
            "Train accuracy: 97.5\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  183\n",
            "Train accuracy: 97.08333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  184\n",
            "Train accuracy: 97.29166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  185\n",
            "Train accuracy: 95.625\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  66.6667\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  186\n",
            "Train accuracy: 97.5\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  187\n",
            "Train accuracy: 96.45833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  188\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  68.3333\n",
            "Epoch:  189\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  190\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  191\n",
            "Train accuracy: 97.08333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  192\n",
            "Train accuracy: 96.45833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  193\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  194\n",
            "Train accuracy: 97.70833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  195\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  196\n",
            "Train accuracy: 97.08333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  197\n",
            "Train accuracy: 96.45833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  198\n",
            "Train accuracy: 96.875\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  199\n",
            "Train accuracy: 97.08333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  200\n",
            "Train accuracy: 97.08333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  201\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  202\n",
            "Train accuracy: 97.70833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  203\n",
            "Train accuracy: 96.66666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  204\n",
            "Train accuracy: 97.70833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  205\n",
            "Train accuracy: 94.58333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  206\n",
            "Train accuracy: 96.45833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  207\n",
            "Train accuracy: 97.70833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  208\n",
            "Train accuracy: 96.04166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  66.6667\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  209\n",
            "Train accuracy: 97.08333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  68.3333\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  210\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  68.3333\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  211\n",
            "Train accuracy: 97.70833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  68.3333\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  212\n",
            "Train accuracy: 96.875\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  68.3333\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  213\n",
            "Train accuracy: 97.70833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  68.3333\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  214\n",
            "Train accuracy: 97.08333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  68.3333\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  215\n",
            "Train accuracy: 97.29166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  68.3333\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  216\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  68.3333\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  217\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  68.3333\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  218\n",
            "Train accuracy: 96.45833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  68.3333\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  219\n",
            "Train accuracy: 96.875\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  68.3333\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  220\n",
            "Train accuracy: 97.5\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  68.3333\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  221\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  68.3333\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  222\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  68.3333\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  68.3333\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  223\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  68.3333\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  224\n",
            "Train accuracy: 97.29166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  68.3333\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  225\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  68.3333\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  226\n",
            "Train accuracy: 97.70833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  227\n",
            "Train accuracy: 97.08333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  228\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  229\n",
            "Train accuracy: 97.70833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  230\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  231\n",
            "Train accuracy: 96.45833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  232\n",
            "Train accuracy: 97.70833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  233\n",
            "Train accuracy: 97.70833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  234\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  235\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  236\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  237\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  238\n",
            "Train accuracy: 97.08333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  239\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  240\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  241\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  242\n",
            "Train accuracy: 96.875\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  243\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  244\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  245\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  246\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  247\n",
            "Train accuracy: 97.08333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  248\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  249\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  250\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  251\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  252\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  253\n",
            "Train accuracy: 99.58333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  68.3333\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  254\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  255\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  256\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  257\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  258\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  259\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  260\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  261\n",
            "Train accuracy: 96.04166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  262\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  263\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  264\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  265\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  266\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  267\n",
            "Train accuracy: 100.0\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  268\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  269\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  270\n",
            "Train accuracy: 97.08333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  271\n",
            "Train accuracy: 96.66666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  272\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  273\n",
            "Train accuracy: 97.70833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  274\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  275\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  276\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  277\n",
            "Train accuracy: 97.70833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  278\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  279\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  70.0\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  280\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  281\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  282\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  283\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  284\n",
            "Train accuracy: 97.5\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  285\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  286\n",
            "Train accuracy: 96.875\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  287\n",
            "Train accuracy: 96.875\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  70.0\n",
            "Epoch:  288\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  289\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  290\n",
            "Train accuracy: 97.08333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  291\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  292\n",
            "Train accuracy: 96.66666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  293\n",
            "Train accuracy: 97.29166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  294\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  295\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  296\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  70.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  297\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  298\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  299\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  300\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  301\n",
            "Train accuracy: 99.58333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  302\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  303\n",
            "Train accuracy: 99.58333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  304\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  305\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  306\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  307\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  308\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  309\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  310\n",
            "Train accuracy: 97.29166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  311\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  312\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  313\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  314\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  315\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  316\n",
            "Train accuracy: 97.29166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  317\n",
            "Train accuracy: 97.70833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  318\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  319\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  320\n",
            "Train accuracy: 99.58333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  321\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  71.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  322\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  323\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  324\n",
            "Train accuracy: 99.79166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  325\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  326\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  71.6667\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  327\n",
            "Train accuracy: 97.70833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  328\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  329\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  330\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  331\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  332\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  333\n",
            "Train accuracy: 97.29166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  334\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  335\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  336\n",
            "Train accuracy: 97.5\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  337\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  338\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  339\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  340\n",
            "Train accuracy: 96.875\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  341\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  342\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  343\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  344\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  345\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  346\n",
            "Train accuracy: 96.875\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  347\n",
            "Train accuracy: 97.5\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  348\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  349\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  75.0\n",
            "Best accuracy with 2.0 perturbation:  73.3333\n",
            "Epoch:  350\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  351\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  352\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  353\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  354\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  355\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  356\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  357\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  358\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  359\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  360\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  361\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  362\n",
            "Train accuracy: 97.5\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  363\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  364\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  365\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  366\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  367\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  368\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  369\n",
            "Train accuracy: 99.58333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  370\n",
            "Train accuracy: 99.58333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  371\n",
            "Train accuracy: 99.58333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  372\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  373\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  374\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  375\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  376\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  377\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  378\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  379\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  380\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  381\n",
            "Train accuracy: 97.29166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  382\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  383\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  384\n",
            "Train accuracy: 99.79166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  385\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  386\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  387\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  388\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  389\n",
            "Train accuracy: 99.58333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  390\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  391\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  392\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  393\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  394\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  395\n",
            "Train accuracy: 97.08333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  396\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  397\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  398\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  399\n",
            "Train accuracy: 97.29166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  400\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  401\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  402\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  403\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  404\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  405\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  406\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  407\n",
            "Train accuracy: 99.58333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  408\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  409\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  410\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  411\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  412\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  413\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  414\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  415\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  416\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  417\n",
            "Train accuracy: 97.5\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  418\n",
            "Train accuracy: 97.5\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  419\n",
            "Train accuracy: 96.875\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  420\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  421\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  422\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  423\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  424\n",
            "Train accuracy: 97.5\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  425\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  426\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  427\n",
            "Train accuracy: 99.58333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  428\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  429\n",
            "Train accuracy: 99.58333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  430\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  431\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  432\n",
            "Train accuracy: 99.58333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  433\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  434\n",
            "Train accuracy: 98.125\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  435\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  436\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  437\n",
            "Train accuracy: 100.0\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  438\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  439\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  440\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  441\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  442\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  443\n",
            "Train accuracy: 99.79166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  444\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  445\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  446\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  447\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  73.3333\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  448\n",
            "Train accuracy: 99.58333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  449\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  450\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  451\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  452\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  453\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  454\n",
            "Train accuracy: 97.91666666666666\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  455\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  456\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  457\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  458\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  459\n",
            "Train accuracy: 99.58333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  460\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  461\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  462\n",
            "Train accuracy: 98.95833333333334\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  463\n",
            "Train accuracy: 99.79166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  464\n",
            "Train accuracy: 99.79166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  465\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  466\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  467\n",
            "Train accuracy: 97.70833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  468\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  469\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  470\n",
            "Train accuracy: 99.16666666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  471\n",
            "Train accuracy: 99.375\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  472\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  473\n",
            "Train accuracy: 99.58333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  474\n",
            "Train accuracy: 97.70833333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  475\n",
            "Train accuracy: 98.75\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  476\n",
            "Train accuracy: 98.54166666666667\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  477\n",
            "Train accuracy: 98.33333333333333\n",
            "Best accuracy without perturbation:  75.0\n",
            "Best accuracy with 0.5 perturbation:  76.6667\n",
            "Best accuracy with 1.0 perturbation:  75.0\n",
            "Best accuracy with 1.5 perturbation:  76.6667\n",
            "Best accuracy with 2.0 perturbation:  76.6667\n",
            "Epoch:  478\n",
            "Train accuracy: 98.75\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9xE_NClW5umI"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ChOPNYKns_9V"
      },
      "source": [
        "# Feature Attacks"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tpY_cL9ds_hC"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IQNFtKHEgRRE"
      },
      "source": [
        "import torch\n",
        "\n",
        "m = Normal(torch.tensor([[0.0, 1.0], [1.0, 0.0]]), torch.tensor([[1.0, 0.5],[0.5, 1.0]]))\n",
        "print(m.sample())"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8_It0ktQX_xj"
      },
      "source": [
        "a = torch.rand(4, 5)\n",
        "print(a)\n",
        "b = torch.rand(3, 5)\n",
        "print(b)\n",
        "c = torch.cat([a, b], dim=0)\n",
        "print(c)"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}