add simple example for simple boosting test

example/mnist_mlp_simple_boosting.cc

@@ -0,0 +1,45 @@
+#include "galois/models.h"
+#include "galois/filters.h"
+#include "galois/dataset/mnist.h"
+
+using namespace std;
+using namespace gs;
+
+int main()
+{
+    using T = float;
+
+    auto path1 = make_shared<Path<T>>();
+    path1->add_filter(make_shared<Linear<T>>(28*28, 512));
+    path1->add_filter(make_shared<Tanh<T>>());
+    path1->add_filter(make_shared<Linear<T>>(512, 10));
+
+    int batch_size = 10;
+    int num_epoch = 5;
+    T learning_rate = 0.05;
+    auto train_images = mnist::read_images<T>("./data/train-images-idx3-ubyte.gz");
+    auto train_labels = mnist::read_labels<T>("./data/train-labels-idx1-ubyte.gz");
+    auto test_images = mnist::read_images<T>("./data/t10k-images-idx3-ubyte.gz");
+    auto test_labels = mnist::read_labels<T>("./data/t10k-labels-idx1-ubyte.gz");
+
+    Model<T> model1(batch_size, num_epoch, learning_rate, "sgd");
+    model1.add_link("images", "output", path1);
+    model1.add_link("output", "predicitons", make_shared<CrossEntropy<T>>());
+    model1.add_input_ids("images");
+    model1.add_output_ids("predicitons");
+    model1.add_train_dataset(train_images, train_labels);
+    model1.add_test_dataset(test_images, test_labels);
+    model1.fit();
+
+    auto path2 = path1->clone();
+    path1->fix_params();
+    Model<T> model2(batch_size, 10, 0.01, "sgd");
+    model2.add_link("images", "output", path1);
+    model2.add_link("images", "output", path2);
+    model2.add_link("output", "predicitons", make_shared<CrossEntropy<T>>());
+    model2.add_input_ids("images");
+    model2.add_output_ids("predicitons");
+    model2.add_train_dataset(train_images, train_labels);
+    model2.add_test_dataset(test_images, test_labels);
+    model2.fit();
+}