network: implemented set_optimizer

examples/mnist/main.cpp

@@ -67,7 +67,10 @@ int main()
     // No. of trainable params
     network.summary();
     network.loss_tol = 2e-2;
-    network.fit( x_train, y_train, 1000, 0.01, true );
+
+    auto opt = Robbie::Optimizers::StochasticGradientDescent( 0.01 );
+    network.set_optimizer( &opt );
+    network.fit( x_train, y_train, 1000, true );
 
     fmt::print( "Loss on test set = {:.3e}\n", network.loss( x_test, y_test ) );
 

include/network.hpp

@@ -9,6 +9,7 @@
 #include <cstddef>
 #include <memory>
 #include <optional>
+#include <stdexcept>
 #include <type_traits>
 #include <vector>
 
@@ -22,9 +23,6 @@ class Network
 public:
     std::optional<scalar> loss_tol = std::nullopt;
 
-    std::unique_ptr<Optimizers::Optimizer<scalar>> opt
-        = std::make_unique<Optimizers::StochasticGradientDescent<scalar>>( 0.00001 );
-
     Network() = default;
 
     template<typename LayerT, typename... T>
@@ -67,10 +65,10 @@ class Network
         return loss;
     }
 
-    // void set_optimizer( const Optimizers::Optimizer<scalar> & opt )
-    // {
-    //     this->opt = s
-    // }
+    void set_optimizer( Optimizers::Optimizer<scalar> * opt )
+    {
+        this->opt = opt;
+    }
 
     void register_optimizer_variables()
     {
@@ -93,8 +91,12 @@ class Network
 
     void
     fit( const std::vector<Matrix<scalar>> & x_train, const std::vector<Matrix<scalar>> & y_train, size_t epochs,
-         scalar learning_rate, bool print_progress = false )
+         bool print_progress = false )
     {
+
+        if( this->opt == nullptr )
+            throw std::runtime_error( "Optimizer has not been set!" );
+
         register_optimizer_variables();
 
         auto n_samples  = x_train.size();
@@ -194,6 +196,7 @@ class Network
 
 private:
     std::vector<std::unique_ptr<Layer<scalar>>> layers;
+    Optimizers::Optimizer<scalar> * opt = nullptr;
 };
 
 } // namespace Robbie

include/robbie.hpp

@@ -6,4 +6,5 @@
 #include "fc_layer.hpp"
 #include "layer.hpp"
 #include "loss_functions.hpp"
-#include "network.hpp"
+#include "network.hpp"
+#include "optimizers.hpp"

main.cpp

@@ -48,7 +48,10 @@ int main()
     fmt::print( "x_train[10] = {}\n", fmt::streamed( x_train[10] ) );
     fmt::print( "y_train[10] = {}\n", fmt::streamed( y_train[10] ) );
 
+    auto opt = Optimizers::StochasticGradientDescent<scalar>( 0.00001 );
+
     auto network = Network<scalar, LossFunctions::MeanSquareError>();
+    network.set_optimizer( &opt );
     network.add<FCLayer<scalar>>( input_size, 100 );
     network.add<ActivationLayer<scalar, ActivationFunctions::Tanh>>();
     network.add<DropoutLayer<scalar>>( 0.5 );
@@ -57,7 +60,7 @@ int main()
     network.add<FCLayer<scalar>>( 30, 10 );
     network.summary();
 
-    network.fit( x_train, y_train, 300, 0.00001, true );
+    network.fit( x_train, y_train, 300, true );
 
     fmt::print( "Loss on test set = {:.3e}\n", network.loss( x_test, y_test ) );
 }

test/test_xor.cpp

@@ -21,15 +21,18 @@ TEST_CASE( "Test_XOR" )
     y_train[2] << 1;
     y_train[3] << 0;
 
-    auto network     = Network<double, LossFunctions::MeanSquareError>();
+    auto opt     = Optimizers::StochasticGradientDescent<double>( 0.01 );
+    auto network = Network<double, LossFunctions::MeanSquareError>();
+    network.set_optimizer( &opt );
+
     network.loss_tol = 5e-4;
 
     network.add<FCLayer<double>>( 2, 10 );
     network.add<ActivationLayer<double, ActivationFunctions::Tanh>>();
     network.add<FCLayer<double>>( 10, 1 );
     network.add<ActivationLayer<double, ActivationFunctions::Tanh>>();
 
-    network.fit( x_train, y_train, 50000, 0.01 );
+    network.fit( x_train, y_train, 50000 );
 
     auto out = network.predict( x_train );