Feature Request: Interactive ANN Playground for Custom Training & Visualization

[HARSHITA005-GARG]

Hii @manasvi-0, I would like to work on the following proposal. Would love to hear it from you soon.

🚀 Feature Proposal: Interactive ANN Playground for Custom Training & Visualization

📌 Summary

I propose adding an Interactive ANN Playground to AlgoLab, allowing users to:

• Upload their own dataset (CSV or from Kaggle API)
• Select Regression or Classification mode
• Customize neural network architecture & training settings
• Visualize what happens during training and inside hidden layers

This feature will make AlgoLab a hands-on deep learning lab for both beginners and advanced users.

---

🎯 Motivation

Currently, AlgoLab demonstrates algorithms with fixed configurations.
However, deep learning is all about experimentation with architecture and hyperparameters.
By creating an interactive ANN module:

• Users can learn the effect of neurons, layers, activation functions, dropout, regularization in real-time
• It bridges theory and practice for those learning deep learning concepts
• It makes AlgoLab dataset-agnostic, as users can bring their own data

---

🛠 Proposed Features

1. Dataset Input

   • Upload CSV file directly
   • Or fetch datasets via Kaggle API
   • Automatic type detection: regression or classification based on target column

2. Model Customization Controls

   • Number of hidden layers
   • Neurons per layer
   • Activation functions (ReLU, Sigmoid, Tanh, Softmax, etc.)
   • Regularization (L1, L2, ElasticNet)
   • Dropout rate
   • Learning rate
   • Optimizer (SGD, Adam, RMSprop, etc.)
   • Batch size
   • Number of epochs

3. Visualization & Insights

   • Live accuracy/loss curves
   • Intermediate hidden layer outputs (activations)
   • Model architecture diagram
   • Weight & bias distribution plots
   • Confusion matrix (classification) or residual plot (regression)

4. Training & Evaluation

   • Real-time training updates
   • Train/validation/test split configuration
   • Performance metrics:
     • Classification → Accuracy, Precision, Recall, F1-score
     • Regression → RMSE, MAE, R²

5. Integration in AlgoLab

   • New section under “Deep Learning” → Interactive ANN Playground
   • Streamlit-based UI for easy interactivity

---

📊 Example Workflow

1. User uploads a CSV file with their dataset.
2. AlgoLab detects the task type (regression/classification).
3. User selects:
   • 3 hidden layers with [64, 32, 16] neurons
   • ReLU activation
   • Dropout 0.2
   • Learning rate 0.001
   • Optimizer = Adam
4. AlgoLab:
   • Trains the model
   • Shows real-time accuracy/loss graphs
   • Visualizes hidden layer activations
   • Displays final evaluation metrics

---

✅ Benefits

• Makes AlgoLab highly interactive & educational
• Allows experimentation without coding
• Suitable for students, educators, and ML enthusiasts
• Encourages hands-on learning of hyperparameter tuning
• Supports any dataset from CSV or Kaggle

---

This ANN Playground will transform AlgoLab into a practical deep learning learning hub.

[manasvi-0]

@HARSHITA005-GARG yes you can work on this after compeleting your task please raise a PR and dont merge it until i review it
and keep the same names as guided