neural_layer_connected.h File Reference

An implementation of a fully-connected layer of perceptrons. More...

#include "neural_layer.h"

Include dependency graph for neural_layer_connected.h:

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

Go to the source code of this file.

Functions
void	neural_layer_connected_init (struct Layer *l, const struct ArgsLayer *args)
	Initialises a fully-connected layer. More...
struct Layer *	neural_layer_connected_copy (const struct Layer *src)
	Initialises and creates a copy of one connected layer from another. More...
void	neural_layer_connected_rand (struct Layer *l)
	Randomises a connected layer weights. More...
void	neural_layer_connected_forward (const struct Layer *l, const struct Net *net, const double *input)
	Forward propagates a connected layer. More...
void	neural_layer_connected_backward (const struct Layer *l, const struct Net *net, const double *input, double *delta)
	Backward propagates a connected layer. More...
void	neural_layer_connected_update (const struct Layer *l)
	Updates the weights and biases of a connected layer. More...
void	neural_layer_connected_print (const struct Layer *l, const bool print_weights)
	Prints a connected layer. More...
bool	neural_layer_connected_mutate (struct Layer *l)
	Mutates a connected layer. More...
void	neural_layer_connected_free (const struct Layer *l)
	Free memory used by a connected layer. More...
double *	neural_layer_connected_output (const struct Layer *l)
	Returns the output from a connected layer. More...
size_t	neural_layer_connected_save (const struct Layer *l, FILE *fp)
	Writes a connected layer to a file. More...
size_t	neural_layer_connected_load (struct Layer *l, FILE *fp)
	Reads a connected layer from a file. More...
void	neural_layer_connected_resize (struct Layer *l, const struct Layer *prev)
	Resizes a connected layer if the previous layer has changed size. More...
char *	neural_layer_connected_json_export (const struct Layer *l, const bool return_weights)
	Returns a json formatted string representation of a connected layer. More...

Variables
static struct LayerVtbl const	layer_connected_vtbl
	Neural connected layer implemented functions. More...

Detailed Description

An implementation of a fully-connected layer of perceptrons.

Author

Richard Preen rpree.nosp@m.n@gm.nosp@m.ail.c.nosp@m.om

Copyright

The Authors.

Date

2016–2021.

Definition in file neural_layer_connected.h.

Function Documentation

neural_layer_connected_backward()

void neural_layer_connected_backward	(	const struct Layer *	l,
		const struct Net *	net,
		const double *	input,
		double *	delta
	)

Backward propagates a connected layer.

Parameters

[in]	l	The layer to backward propagate.
[in]	net	Network containing the layer.
[in]	input	The input to the layer.
[out]	delta	The previous layer's error.

Definition at line 196 of file neural_layer_connected.c.

References Layer::bias_updates, blas_axpy(), blas_gemm(), Layer::c, Layer::delta, Layer::function, LAYER_SGD_WEIGHTS, Layer::n_inputs, Layer::n_outputs, neural_gradient_array(), Layer::options, Layer::state, Layer::weight_updates, and Layer::weights.

Here is the call graph for this function:

neural_layer_connected_copy()

struct Layer* neural_layer_connected_copy

(

const struct Layer *

src

)

Initialises and creates a copy of one connected layer from another.

Parameters

[in]

src

The source layer.

Returns

A pointer to the new layer.

Definition at line 122 of file neural_layer_connected.c.

References Layer::biases, CONNECTED, Layer::decay, Layer::eta, Layer::eta_max, Layer::eta_min, Layer::function, layer_defaults(), Layer::layer_vptr, malloc_layer_arrays(), Layer::max_neuron_grow, Layer::max_outputs, Layer::momentum, Layer::mu, Layer::n_active, Layer::n_biases, Layer::n_inputs, N_MU, Layer::n_outputs, Layer::n_weights, Layer::options, Layer::out_c, Layer::out_h, Layer::out_w, Layer::type, Layer::weight_active, and Layer::weights.

Here is the call graph for this function:

neural_layer_connected_forward()

void neural_layer_connected_forward	(	const struct Layer *	l,
		const struct Net *	net,
		const double *	input
	)

Forward propagates a connected layer.

Parameters

[in]	l	Layer to forward propagate.
[in]	net	Network containing the layer.
[in]	input	Input to the layer.

Definition at line 174 of file neural_layer_connected.c.

References Layer::biases, blas_gemm(), Layer::c, Layer::function, Layer::n_inputs, Layer::n_outputs, neural_activate_array(), Layer::output, Layer::state, and Layer::weights.

Here is the call graph for this function:

neural_layer_connected_free()

void neural_layer_connected_free

(

const struct Layer *

l

)

Free memory used by a connected layer.

Parameters

[in]

l

The layer to be freed.

Definition at line 103 of file neural_layer_connected.c.

References Layer::bias_updates, Layer::biases, Layer::delta, Layer::mu, Layer::output, Layer::state, Layer::weight_active, Layer::weight_updates, and Layer::weights.

neural_layer_connected_init()

void neural_layer_connected_init	(	struct Layer *	l,
		const struct ArgsLayer *	args
	)

Initialises a fully-connected layer.

Parameters

[in]	l	Layer to initialise.
[in]	args	Parameters to initialise the layer.

Definition at line 70 of file neural_layer_connected.c.

References Layer::biases, Layer::decay, ArgsLayer::decay, ArgsLayer::eta, Layer::eta_max, Layer::eta_min, ArgsLayer::eta_min, Layer::function, ArgsLayer::function, layer_args_opt(), layer_init_eta(), malloc_layer_arrays(), Layer::max_neuron_grow, ArgsLayer::max_neuron_grow, Layer::max_outputs, Layer::momentum, ArgsLayer::momentum, Layer::mu, MU_TYPE, Layer::n_active, Layer::n_biases, ArgsLayer::n_init, Layer::n_inputs, ArgsLayer::n_inputs, ArgsLayer::n_max, N_MU, Layer::n_outputs, Layer::n_weights, Layer::options, Layer::out_c, Layer::out_h, Layer::out_w, rand_normal(), sam_init(), Layer::weight_active, WEIGHT_SD_INIT, and Layer::weights.

Here is the call graph for this function:

neural_layer_connected_json_export()

char* neural_layer_connected_json_export	(	const struct Layer *	l,
		const bool	return_weights
	)

Returns a json formatted string representation of a connected layer.

Parameters

[in]	l	The layer to return.
[in]	return_weights	Whether to returnprint the values of weights and biases.

Returns

String encoded in json format.

Definition at line 354 of file neural_layer_connected.c.

References Layer::eta, Layer::function, layer_weight_json(), Layer::mu, Layer::n_inputs, N_MU, Layer::n_outputs, neural_activation_string(), and Layer::weights.

Referenced by neural_layer_connected_print().

Here is the call graph for this function:

Here is the caller graph for this function:

neural_layer_connected_load()

size_t neural_layer_connected_load	(	struct Layer *	l,
		FILE *	fp
	)

Reads a connected layer from a file.

Parameters

[in]	l	The layer to load.
[in]	fp	Pointer to the file to be read.

Returns

The number of elements read.

Definition at line 415 of file neural_layer_connected.c.

References Layer::bias_updates, Layer::biases, Layer::decay, Layer::eta, Layer::eta_max, Layer::eta_min, Layer::function, malloc_layer_arrays(), Layer::max_neuron_grow, Layer::max_outputs, Layer::momentum, Layer::mu, Layer::n_active, Layer::n_biases, Layer::n_inputs, N_MU, Layer::n_outputs, Layer::n_weights, Layer::options, Layer::out_c, Layer::out_h, Layer::out_w, Layer::weight_active, Layer::weight_updates, and Layer::weights.

Here is the call graph for this function:

neural_layer_connected_mutate()

bool neural_layer_connected_mutate

(

struct Layer *

l

)

Mutates a connected layer.

Parameters

[in]

l

The layer to mutate.

Returns

Whether any alterations were made.

Definition at line 292 of file neural_layer_connected.c.

References layer_add_neurons(), layer_ensure_input_represention(), LAYER_EVOLVE_CONNECT, LAYER_EVOLVE_ETA, LAYER_EVOLVE_FUNCTIONS, LAYER_EVOLVE_NEURONS, LAYER_EVOLVE_WEIGHTS, layer_mutate_connectivity(), layer_mutate_eta(), layer_mutate_functions(), layer_mutate_neurons(), layer_mutate_weights(), Layer::mu, MU_TYPE, N_MU, Layer::options, and sam_adapt().

Here is the call graph for this function:

neural_layer_connected_output()

double* neural_layer_connected_output

(

const struct Layer *

l

)

Returns the output from a connected layer.

Parameters

[in]

l

The layer whose output to return.

Returns

The layer output.

Definition at line 328 of file neural_layer_connected.c.

References Layer::output.

neural_layer_connected_print()

void neural_layer_connected_print	(	const struct Layer *	l,
		const bool	print_weights
	)

Prints a connected layer.

Parameters

[in]	l	The layer to print.
[in]	print_weights	Whether to print the values of weights and biases.

Definition at line 339 of file neural_layer_connected.c.

References neural_layer_connected_json_export().

Here is the call graph for this function:

neural_layer_connected_rand()

void neural_layer_connected_rand

(

struct Layer *

l

)

Randomises a connected layer weights.

Parameters

[in]

l

The layer to randomise.

Definition at line 162 of file neural_layer_connected.c.

References layer_weight_rand().

Here is the call graph for this function:

neural_layer_connected_resize()

void neural_layer_connected_resize	(	struct Layer *	l,
		const struct Layer *	prev
	)

Resizes a connected layer if the previous layer has changed size.

Parameters

[in]	l	The layer to resize.
[in]	prev	The layer previous to the one being resized.

Definition at line 246 of file neural_layer_connected.c.

References Layer::i, layer_calc_n_active(), layer_ensure_input_represention(), LAYER_EVOLVE_CONNECT, layer_print(), Layer::n_inputs, Layer::n_outputs, Layer::n_weights, N_WEIGHTS_MAX, Layer::options, rand_normal(), Layer::weight_active, WEIGHT_SD, Layer::weight_updates, and Layer::weights.

Here is the call graph for this function:

neural_layer_connected_save()

size_t neural_layer_connected_save	(	const struct Layer *	l,
		FILE *	fp
	)

Writes a connected layer to a file.

Parameters

[in]	l	The layer to save.
[in]	fp	Pointer to the file to be written.

Returns

The number of elements written.

Definition at line 382 of file neural_layer_connected.c.

References Layer::bias_updates, Layer::biases, Layer::decay, Layer::eta, Layer::eta_max, Layer::eta_min, Layer::function, Layer::max_neuron_grow, Layer::max_outputs, Layer::momentum, Layer::mu, Layer::n_active, Layer::n_biases, Layer::n_inputs, N_MU, Layer::n_outputs, Layer::n_weights, Layer::options, Layer::weight_active, Layer::weight_updates, and Layer::weights.

neural_layer_connected_update()

void neural_layer_connected_update

(

const struct Layer *

l

)

Updates the weights and biases of a connected layer.

Parameters

[in]

l

The layer to update.

Definition at line 225 of file neural_layer_connected.c.

References Layer::bias_updates, Layer::biases, blas_axpy(), blas_scal(), Layer::decay, Layer::eta, LAYER_SGD_WEIGHTS, layer_weight_clamp(), Layer::momentum, Layer::n_biases, Layer::n_weights, Layer::options, Layer::weight_updates, and Layer::weights.

Here is the call graph for this function:

Variable Documentation

layer_connected_vtbl

struct LayerVtbl const layer_connected_vtbl

static

Initial value:

= {
    &neural_layer_connected_init,     &neural_layer_connected_mutate,
    &neural_layer_connected_resize,   &neural_layer_connected_copy,
    &neural_layer_connected_free,     &neural_layer_connected_rand,
    &neural_layer_connected_print,    &neural_layer_connected_update,
    &neural_layer_connected_backward, &neural_layer_connected_forward,
    &neural_layer_connected_output,   &neural_layer_connected_save,
    &neural_layer_connected_load,     &neural_layer_connected_json_export
}

Neural connected layer implemented functions.

Definition at line 70 of file neural_layer_connected.h.

Referenced by layer_set_vptr().