Neural network functionality for a 3-layer (INPUT, HIDDEN, OUTPUT) feed-forward, back-prop NN. More...

#include <stdlib.h>
#include <string.h>
#include <math.h>
#include "util/mnist-utils.h"
#include "3lnn.h"

Functions
Node *	getNode (Layer *l, int nodeId)

Layer *	getLayer (Network *nn, LayerType ltype)
	Returns one of the layers of the network. More...

double	getActFctDerivative (Network *nn, LayerType ltype, double outVal)
	Returns the result of applying the given outputValue to the derivate of the activation function. More...

void	updateNodeWeights (Network *nn, LayerType ltype, int id, double error)
	Updates a node's weights based on given error. More...

void	backPropagateHiddenLayer (Network *nn, int targetClassification)
	Back propagates network error to hidden layer. More...

void	backPropagateOutputLayer (Network *nn, int targetClassification)
	Back propagates network error in output layer. More...

void	backPropagateNetwork (Network *nn, int targetClassification)
	Back propagates network error from output layer to hidden layer. More...

void	activateNode (Network *nn, LayerType ltype, int id)
	Performs an activiation function (as defined in the NN's defaults) to a specified node. More...

void	calcNodeOutput (Network *nn, LayerType ltype, int id)
	Calculates the output value of a specified node by multiplying all its weights with the previous layer's outputs. More...

void	calcLayer (Network *nn, LayerType ltype)
	Calculates the output values of a given NN layer. More...

void	feedForwardNetwork (Network *nn)
	Feeds input layer values forward to hidden to output layer (calculation and activation fct) More...

void	feedInput (Network nn, Vector v)
	Feeds some Vector data into the INPUT layer of the NN. More...

Layer *	createInputLayer (int inpCount)

Layer *	createLayer (int nodeCount, int weightCount)

void	initNetwork (Network *nn, int inpCount, int hidCount, int outCount)
	Initializes the NN by creating and copying INTPUT, HIDDEN, OUTPUT data structures into the NN's memory space. More...

void	setNetworkDefaults (Network *nn)
	Sets the default network parameters (which can be overwritten/changed) More...

void	initWeights (Network *nn, LayerType ltype)
	Initializes a layer's weights with random values. More...

Network *	createNetwork (int inpCount, int hidCount, int outCount)
	Creates a dynamically-sized, 3-layer (INTPUT, HIDDEN, OUTPUT) neural network. More...

int	getNetworkClassification (Network *nn)
	Returns the network's classification using the ID of teh node with the hightest output. More...

Detailed Description

Neural network functionality for a 3-layer (INPUT, HIDDEN, OUTPUT) feed-forward, back-prop NN.

Author: Matt Lind

Date: August 2015

Definition in file 3lnn.c.

Function Documentation

void activateNode	(	Network *	nn,
		LayerType	ltype,
		int	id
	)

Performs an activiation function (as defined in the NN's defaults) to a specified node.

Parameters

nn	A pointer to the NN
ltype	Type of layer (INPUT, HIDDEN, OUTPUT)
id	Sequential id of the node that is to be calculated

Definition at line 224 of file 3lnn.c.

void backPropagateHiddenLayer	(	Network *	nn,
		int	targetClassification
	)

Back propagates network error to hidden layer.

Parameters

nn	A pointer to the NN
targetClassification	Correct classification (=label) of the input stream

Definition at line 140 of file 3lnn.c.

void backPropagateNetwork	(	Network *	nn,
		int	targetClassification
	)

Back propagates network error from output layer to hidden layer.

Parameters

nn	A pointer to the NN
targetClassification	Correct classification (=label) of the input stream

Definition at line 206 of file 3lnn.c.

void backPropagateOutputLayer	(	Network *	nn,
		int	targetClassification
	)

Back propagates network error in output layer.

Parameters

nn	A pointer to the NN
targetClassification	Correct classification (=label) of the input stream

Definition at line 178 of file 3lnn.c.

void calcLayer	(	Network *	nn,
		LayerType	ltype
	)

Calculates the output values of a given NN layer.

Parameters

nn	A pointer to the NN
ltype	Type of layer (INPUT, HIDDEN, OUTPUT)

Definition at line 288 of file 3lnn.c.

void calcNodeOutput	(	Network *	nn,
		LayerType	ltype,
		int	id
	)

Calculates the output value of a specified node by multiplying all its weights with the previous layer's outputs.

Parameters

nn	A pointer to the NN
ltype	Type of layer (INPUT, HIDDEN, OUTPUT)
id	Sequential id of the node that is to be calculated

Definition at line 249 of file 3lnn.c.

Layer* createInputLayer ( int inpCount )

Creates an input layer and sets all weights to random values [0-1]

Parameters

inpCount Number of nodes in the input layer

Definition at line 344 of file 3lnn.c.

Layer* createLayer	(	int	nodeCount,
		int	weightCount
	)

Creates a layer and sets all weights to random values [0-1]

Parameters

nodeCount	Number of nodes
weightCount	Number of weights per node

Definition at line 378 of file 3lnn.c.

Network* createNetwork	(	int	inpCount,
		int	hidCount,
		int	outCount
	)

Creates a dynamically-sized, 3-layer (INTPUT, HIDDEN, OUTPUT) neural network.

Parameters

inpCount	Number of nodes in the INPUT layer
hidCount	Number of nodes in the HIDDEN layer
outCount	Number of nodes in the OUTPUT layer

Definition at line 505 of file 3lnn.c.

void feedForwardNetwork ( Network * nn )

Feeds input layer values forward to hidden to output layer (calculation and activation fct)

Parameters

nn	A pointer to the NN

Definition at line 306 of file 3lnn.c.

void feedInput	(	Network *	nn,
		Vector *	v
	)

Feeds some Vector data into the INPUT layer of the NN.

Parameters

nn	A pointer to the NN
v	A pointer to a vector

Definition at line 320 of file 3lnn.c.

double getActFctDerivative	(	Network *	nn,
		LayerType	ltype,
		double	outVal
	)

Returns the result of applying the given outputValue to the derivate of the activation function.

Parameters

nn	A pointer to the NN
ltype	Type of layer (INPUT, HIDDEN, OUTPUT)
outVal	Output value that is to be back propagated

Definition at line 78 of file 3lnn.c.

Layer* getLayer	(	Network *	nn,
		LayerType	ltype
	)

Returns one of the layers of the network.

Parameters

nn	A pointer to the NN
ltype	Type of layer to be returned (INPUT, HIDDEN, OUTPUT)

Definition at line 41 of file 3lnn.c.

int getNetworkClassification ( Network * nn )

Returns the network's classification using the ID of teh node with the hightest output.

Parameters

nn	A pointer to the NN

Definition at line 553 of file 3lnn.c.

Node* getNode	(	Layer *	l,
		int	nodeId
	)

Retrieves a node via ID from a layer

Definition at line 22 of file 3lnn.c.

void initNetwork	(	Network *	nn,
		int	inpCount,
		int	hidCount,
		int	outCount
	)

Initializes the NN by creating and copying INTPUT, HIDDEN, OUTPUT data structures into the NN's memory space.

Parameters

nn	A pointer to the NN
inpCount	Number of nodes in the INPUT layer
hidCount	Number of nodes in the HIDDEN layer
outCount	Number of nodes in the OUTPUT layer

Definition at line 413 of file 3lnn.c.

void initWeights	(	Network *	nn,
		LayerType	ltype
	)

Initializes a layer's weights with random values.

Parameters

nn	A pointer to the NN
ltype	Defining what layer to initialize

Definition at line 467 of file 3lnn.c.

void setNetworkDefaults ( Network * nn )

Sets the default network parameters (which can be overwritten/changed)

Parameters

nn	A pointer to the NN

Definition at line 447 of file 3lnn.c.

void updateNodeWeights	(	Network *	nn,
		LayerType	ltype,
		int	id,
		double	error
	)

Updates a node's weights based on given error.

Parameters

nn	A pointer to the NN
ltype	Type of layer (INPUT, HIDDEN, OUTPUT)
id	Sequential id of the node that is to be calculated
error	The error (difference between desired output and actual output

Definition at line 103 of file 3lnn.c.

Functions

Detailed Description

Function Documentation