train.h File Reference

PCFNN_TRAIN. More...

#include "../../config.h"
#include <stdlib.h>
#include <math.h>
#include "neuron.h"
#include "layer.h"
#include "network.h"
#include "feedforward.h"
#include "backprop.h"
#include "batch.h"

Include dependency graph for train.h:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Functions
double *	PCFNN_NETWORK_train (struct PCFNN_NETWORK *net, double **data, double **target, size_t size, double validation_split, int shuffle, unsigned long batch_size, size_t epochs, double eta, double alpha, double(*f_cost)(double, double), double(*f_cost_de)(double, double), double *status)
	Train the network net. More...

Detailed Description

PCFNN_TRAIN.

Author

Cedric FARINAZZO

Version

0.1

Date

9 may 2019

Train functions for PCFNN neural network

Definition in file train.h.

Function Documentation

PCFNN_NETWORK_train()

PCFNN_NETWORK_train	(	struct PCFNN_NETWORK *	net,
		double **	data,
		double **	target,
		size_t	size,
		double	validation_split,
		int	shuffle,
		unsigned long	batch_size,
		size_t	epochs,
		double	eta,
		double	alpha,
		double(*)(double, double)	f_cost,
		double(*)(double, double)	f_cost_de,
		double *	status
	)

Train the network net.

Parameters

[in]	net	(struct PCFNN_NETWORK*) a pointer an a PCFNN_NETWORK
[in]	data	(double**) double array array: an array of input data (lenght of data is size)
[in]	target	(double**) double array array : an array of expected output of the output layer of net (lenght of target is size)
[in]	size	(size_t) lenght of data and target
[in]	validation_split	(double) a double between 0 and 1: the part of the dataset to use to test the network
[in]	shuffle	(int) 1 to enable shuffle mode and 0 to disable
[in]	batch_size	(unsigned long) batch size
[in]	epochs	(size_t) number of epochs to run
[in]	eta	(double) learning rate
[in]	alpha	(double) momentum rate
[in]	f_cost	(double(*f_cost)(double, double)) a cost function pointer
[in]	f_cost_de	(double(*f_cost_de)(double, double)) a pointer to the f_cost derivative
[out]	status	(double*) a pointer on a double. Can be NULL. It will contain the percentage of completion of the training. Usefull with thread

Returns

double array of net->outputl->size elements which is the current loss for each output (must be free after usage) if validation_split > 0 else return NULL. NULL if wrong parameters.

Definition at line 26 of file train.c.

Here is the call graph for this function: