fixed indents

doc/ipython-notebooks/neuralnets/neuralnets_digits.ipynb

@@ -87,11 +87,11 @@
        "output_type": "stream",
        "stream": "stdout",
        "text": [
-        "Test Error = 8.58938547486 %\n"
+        "Test Error = 8.54283054004 %\n"
        ]
       }
      ],
-     "prompt_number": 4
+     "prompt_number": 1
     },
     {
      "cell_type": "code",

src/shogun/neuralnets/NeuralLayer.cpp

@@ -85,13 +85,3 @@ void CNeuralLayer::shallow_copy(const CNeuralLayer &orig)
 	m_input_gradients = SGVector<float64_t>(orig.m_input_gradients);
 	m_local_gradients = SGVector<float64_t>(orig.m_local_gradients);
 }
-
-
-
-
-
-
-
-
-
-

src/shogun/neuralnets/NeuralLayer.h

@@ -96,8 +96,8 @@ class CNeuralLayer : public CSGObject
 	 * parameters
 	 */
 	virtual void initialize_parameters(float64_t* parameters,
-										bool* parameter_regularizable,
-										float64_t sigma = 0.01f) = 0;
+			bool* parameter_regularizable,
+			float64_t sigma = 0.01f) = 0;
 
 	/** Computes the activations of the neurons in this layer, results should 
 	 * be stored in m_activations
@@ -136,10 +136,10 @@ class CNeuralLayer : public CSGObject
 	 * layer
 	 */
 	virtual void compute_gradients(float64_t* parameters, 
-									bool is_output,
-									float64_t* p,
-									float64_t* previous_layer_activations,
-									float64_t* parameter_gradients) = 0;
+			bool is_output,
+			float64_t* p,
+			float64_t* previous_layer_activations,
+			float64_t* parameter_gradients) = 0;
 
 	/** Computes the error between the layer's current activations and the given
 	 * target activations. Should only be used with output layers

src/shogun/neuralnets/NeuralLinearLayer.h

@@ -64,8 +64,8 @@ class CNeuralLinearLayer : public CNeuralLayer
 	 * parameters
 	 */
 	virtual void initialize_parameters(float64_t* parameters,
-										bool* parameter_regularizable,
-										float64_t sigma = 0.01f);
+			bool* parameter_regularizable,
+			float64_t sigma = 0.01f);
 
 	/** Computes the activations of the neurons in this layer, results should 
 	 * be stored in m_activations
@@ -77,7 +77,7 @@ class CNeuralLinearLayer : public CNeuralLayer
 	 * previous layer, matrix of size previous_layer_num_neurons * batch_size
 	 */
 	virtual void compute_activations(float64_t* parameters,
-									float64_t* previous_layer_activations);
+			float64_t* previous_layer_activations);
 
 	/** Computes the gradients that are relevent to this layer:
 	 *		- The gradients of the error with respect to the layer's parameters
@@ -104,10 +104,10 @@ class CNeuralLinearLayer : public CNeuralLayer
 	 * layer
 	 */
 	virtual void compute_gradients(float64_t* parameters, 
-									bool is_output,
-									float64_t* p,
-									float64_t* previous_layer_activations,
-									float64_t* parameter_gradients);
+			bool is_output,
+			float64_t* p,
+			float64_t* previous_layer_activations,
+			float64_t* parameter_gradients);
 
 	/** Computes the error between the layer's current activations and the given
 	 * target activations. Should only be used with output layers
@@ -144,4 +144,4 @@ class CNeuralLinearLayer : public CNeuralLayer
 };
 
 }
-#endif
+#endif

src/shogun/neuralnets/NeuralLogisticLayer.cpp

@@ -29,7 +29,7 @@ CNeuralLinearLayer(num_neurons)
 }
 
 void CNeuralLogisticLayer::compute_activations(float64_t* parameters,
-										float64_t* previous_layer_activations)
+		float64_t* previous_layer_activations)
 {
 	CNeuralLinearLayer::compute_activations(parameters, 
 											previous_layer_activations);
@@ -41,7 +41,7 @@ void CNeuralLogisticLayer::compute_activations(float64_t* parameters,
 }
 
 void CNeuralLogisticLayer::compute_local_gradients(bool is_output, 
-														float64_t* p)
+		float64_t* p)
 {
 	CNeuralLinearLayer::compute_local_gradients(is_output,p);
 
@@ -50,13 +50,3 @@ void CNeuralLogisticLayer::compute_local_gradients(bool is_output,
 	for (int32_t i=0; i<length; i++)
 		m_local_gradients[i] *= m_activations[i] * (1.0-m_activations[i]);
 }
-
-
-
-
-
-
-
-
-
-

src/shogun/neuralnets/NeuralLogisticLayer.h

@@ -10,8 +10,6 @@
 #ifndef __NEURALLOGISTICLAYER_H__
 #define __NEURALLOGISTICLAYER_H__
 
-#include <shogun/lib/config.h>
-#include <shogun/lib/common.h>
 #include <shogun/neuralnets/NeuralLinearLayer.h>
 
 #ifdef HAVE_EIGEN3
@@ -52,7 +50,7 @@ class CNeuralLogisticLayer : public CNeuralLinearLayer
 	 * previous layer, matrix of size previous_layer_num_neurons * batch_size
 	 */
 	virtual void compute_activations(float64_t* parameters,
-									float64_t* previous_layer_activations);
+			float64_t* previous_layer_activations);
 
 	/** Computes the gradients of the error with respect to this layer's
 	 * activations. Results are stored in m_local_gradients. 
@@ -76,4 +74,4 @@ class CNeuralLogisticLayer : public CNeuralLinearLayer
 };
 
 }
-#endif
+#endif

src/shogun/neuralnets/NeuralNetwork.cpp

@@ -76,7 +76,7 @@ CNeuralNetwork::~CNeuralNetwork()
 }
 
 CDenseFeatures<float64_t>* CNeuralNetwork::apply(
-											CDenseFeatures<float64_t>* inputs)
+		CDenseFeatures<float64_t>* inputs)
 {
 	ASSERT(inputs->get_num_features()==m_num_inputs);
 
@@ -101,12 +101,12 @@ CDenseFeatures<float64_t>* CNeuralNetwork::apply(
 }
 
 void CNeuralNetwork::train_gradient_descent(
-										CDenseFeatures< float64_t >* inputs,	
-										CDenseFeatures< float64_t >* targets,
-										int32_t max_num_epochs,
-										int32_t batch_size, 
-										float64_t learning_rate,		
-										float64_t momentum)
+		CDenseFeatures< float64_t >* inputs,	
+		CDenseFeatures< float64_t >* targets,
+		int32_t max_num_epochs,
+		int32_t batch_size, 
+		float64_t learning_rate,		
+		float64_t momentum)
 {
 	int32_t training_set_size = inputs->get_num_vectors();
 	int32_t _batch_size = batch_size;
@@ -321,20 +321,3 @@ void CNeuralNetwork::shallow_copy(const CNeuralNetwork& orig)
 	m_param_gradients = SGVector<float64_t>(orig.m_param_gradients);
 	m_param_regularizable = SGVector<bool>(orig.m_param_regularizable);
 }
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-

src/shogun/neuralnets/NeuralNetwork.h

@@ -91,11 +91,11 @@ class CNeuralNetwork : public CSGObject
 	 * @param momentum momentum multiplier
 	 */
 	virtual void train_gradient_descent(CDenseFeatures<float64_t>* inputs,
-										CDenseFeatures<float64_t>* targets,
-										int32_t max_num_epochs = 1000,
-										int32_t batch_size = 0,
-										float64_t learning_rate = 0.1,
-										float64_t momentum = 0.9);
+			CDenseFeatures<float64_t>* targets,
+			int32_t max_num_epochs = 1000,
+			int32_t batch_size = 0,
+			float64_t learning_rate = 0.1,
+			float64_t momentum = 0.9);
 
 	/** Checks if the gradients computed using backpropagation are correct by 
 	 * comparing them with gradients computed using numerical approximation.
@@ -109,7 +109,7 @@ class CNeuralNetwork : public CSGObject
 	 * @return true if the gradients are correct, false otherwise
 	 */
 	virtual bool check_gradients(float64_t epsilon=1.0e-06, 
-								 float64_t tolerance=1.0e-09);
+			float64_t tolerance=1.0e-09);
 
 	/** returns the totat number of parameters in the network */
 	int32_t get_num_parameters() {return m_total_num_parameters;}
@@ -177,7 +177,7 @@ class CNeuralNetwork : public CSGObject
 	 * update the activations before the error is computed.
 	 */
 	virtual float64_t compute_error(float64_t* targets, 
-									float64_t* inputs=NULL);
+			float64_t* inputs=NULL);
 
 private:
 	/** returns a pointer to layer i in the network */
@@ -254,4 +254,4 @@ class CNeuralNetwork : public CSGObject
 };
 
 }
-#endif
+#endif