Index - Fast Artificial Neural Network Library (FANN)

I
	Implementation
	init_weights, neural_net
fann_	init_weights
	Installation Using PEAR
	Installing FANN
K
	Known Problems
L
	LAYER
	length_train_data, training_data
fann_	length_train_data
FANN_	LINEAR fann_cpp.h fann_data.h
FANN_	LINEAR_PIECE fann_cpp.h fann_data.h
FANN_	LINEAR_PIECE_SYMMETRIC fann_cpp.h fann_data.h
M
	MAKE_NAME(compute_MSE)
	MAKE_NAME(layer_constructor) fann_generic.h optimized/ fann_optimized_template.h
	MAKE_NAME(layer_destructor)
	MAKE_NAME(layer_run) fann_generic.h optimized/ fann_optimized_template.h
	MAKE_NAME(layer_train_initialize) fann_generic.h optimized/ fann_optimized_template.h
	MAKE_NAME(neuron_backprop) fann_generic.h optimized/ fann_optimized_template.h
	MAKE_NAME(neuron_compute_MSE)
	MAKE_NAME(neuron_constructor) fann_generic.h optimized/ fann_optimized_template.h
	MAKE_NAME(neuron_destructor)
	MAKE_NAME(neuron_private_data) fann_generic.h optimized/ blas/ fann_batch.h optimized/ blas/ fann_rprop.h optimized/ scalar/ fann_batch.h optimized/ scalar/ fann_rprop.h optimized/ sse/ fann_batch.h optimized/ sse/ fann_rprop.h
	MAKE_NAME(neuron_run) fann_generic.h optimized/ fann_optimized_template.h
	MAKE_NAME(neuron_train_initialize) fann_generic.h optimized/ fann_optimized_template.h
	MAKE_NAME(neuron_update) fann_generic.h optimized/ fann_optimized_template.h
	merge_train_data, training_data
fann_	merge_train_data
N
FANN_	NETTYPE_LAYER
FANN_	NETTYPE_SHORTCUT
	Network Design
	network_type_enum, FANN
fann_	network_type_enum
FANN_	NETWORK_TYPE_NAMES
	Neural Network Theory
	Neural Networks
	Neural Networks on the GPU
	neural_net Global neural_net
~	neural_net, neural_net
	Notes and differences from C API
	num_input_train_data, training_data
fann_	num_input_train_data
	num_output_train_data, training_data
fann_	num_output_train_data
P
	Parameters fann.h fann_cascade.h fann_train.h
	PHP Extension
	Precision of a Fixed Point ANN
	print_connections, neural_net
fann_	print_connections
	print_error, neural_net
fann_	print_error
	print_parameters, neural_net
fann_	print_parameters
	Python Bindings
	Python Install
R
	randomize_weights, neural_net
fann_	randomize_weights
	read_train_from_file, training_data
fann_	read_train_from_file
	Requirements
	reset_errno, neural_net
fann_	reset_errno
	reset_errstr, neural_net
fann_	reset_errstr
	reset_MSE, neural_net
fann_	reset_MSE
	RPM Installation
	run, neural_net
fann_	run
	Running a Fixed Point ANN

Each Layer struct holds all variables such as input vector and weight matrix that are represented as textures.

void init_weights( const training_data & data )

Initialize the weights using Widrow + Nguyen’s algorithm.

FANN_EXTERNAL void FANN_API fann_init_weights(
struct fann * ann,
struct fann_train_data * train_data
)

Initialize the weights using Widrow + Nguyen’s algorithm.

The easiest way to install FANN-PHP is to use PEAR- if you have a fairly recent version of PHP installed, simply run pear install fann.

Copies of FANN can be obtained from our SourceForge page, located at http://fann.sf.net

If you are getting in trouble to use your own compiled FANN DLL with Delphi that might be because of the C++ naming mangle that changes between C++ compilers.

Each layer only has connections to the next layer

unsigned int length_train_data()

Returns the number of training patterns in the training_data.

FANN_EXTERNAL unsigned int FANN_API fann_length_train_data(
struct fann_train_data * data
)

Returns the number of training patterns in the struct fann_train_data.

Linear activation function.

Bounded linear activation function.

Bounded Linear activation function.

FANN_EXTERNAL void FANN_API MAKE_NAME(
neuron_compute_MSE
)(struct fann *ann, struct fann_neuron *neuron, fann_type *desired_output)

Compute the error on a MIMO Neuron after forward propagation of a certain input vector i.e.

FANN_EXTERNAL int FANN_API MAKE_NAME(
layer_constructor
)(struct fann *ann, struct fann_layer *layer, struct fann_layer_descr *descr)

Allocates room inside the layer for neurons and connections.

FANN_EXTERNAL void FANN_API MAKE_NAME(
layer_destructor
)(struct fann_layer* layer)

destroy the layer allocated data

FANN_EXTERNAL void FANN_API MAKE_NAME(
layer_run
)(struct fann *ann, struct fann_layer* layer)

Iterates over all neurons and calls the respective run functions.

FANN_EXTERNAL void FANN_API MAKE_NAME(
layer_train_initialize
)(struct fann *ann, struct fann_layer *layer)

Allocates room inside the layer for training data structures.

FANN_EXTERNAL void FANN_API MAKE_NAME(
neuron_backprop
)(struct fann *ann, struct fann_neuron *neuron, fann_type *prev_layer_errors)

Train the MIMO neuron: this function backpropagates the error to the previous layer and computes the weight update.

FANN_EXTERNAL void FANN_API MAKE_NAME(
neuron_compute_MSE
)(struct fann *ann, struct fann_neuron *neuron, fann_type *desired_output)

Compute the error on a MIMO Neuron after forward propagation of a certain input vector i.e.

FANN_EXTERNAL int FANN_API MAKE_NAME(
neuron_constructor
)(struct fann *ann, struct fann_layer *layer, struct fann_neuron *neuron, struct fann_neuron_descr * descr)

Allocates room inside the neuron for the connections.

FANN_EXTERNAL void FANN_API MAKE_NAME(
neuron_destructor
)(struct fann_neuron* neuron)

destroy the neuron allocated data

This structure just stores the arrays needed to use all the algorims supported in old fann code.

FANN_EXTERNAL void FANN_API MAKE_NAME(
neuron_run
)(struct fann *ann, struct fann_neuron *neuron)

Does a forward iteration on the neuron.

FANN_EXTERNAL void FANN_API MAKE_NAME(
neuron_train_initialize
)(struct fann *ann, struct fann_layer *layer, struct fann_neuron *neuron)

Allocates room inside the neuron for training data structures.

FANN_EXTERNAL void FANN_API MAKE_NAME(
neuron_update
)(struct fann *ann, struct fann_neuron *neuron)

Apply the training on the MIMO neuron: the weight update stored in the neuron is applied.

void merge_train_data( const training_data & data )

Merges the data into the data contained in the training_data.

FANN_EXTERNAL struct fann_train_data *FANN_API fann_merge_train_data(
struct fann_train_data * data1,
struct fann_train_data * data2
)

Merges the data from data1 and data2 into a new struct fann_train_data.

Each layer only has connections to the next layer

Each layer has connections to all following layers

When creating a network it is necessary to define how many layers, neurons and connections it should have.

Definition of network types used by neural_net::get_network_type

Definition of network types used by fann_get_network_type

Constant array consisting of the names for the network types, so that the name of an network type can be received by:

This section will briefly explain the theory of neural networks (hereafter known as NN) and artificial neural networks (hereafter known as ANN).

The human brain is a highly complicated machine capable of solving very complex problems.

This section describes the implementation of the current state of the neural networks on the gpu using GLSL project.

Encapsulation of a neural network struct fann and associated C API functions.

neural_net( ) : ann(NULL)

Default constructor creates an empty neural net.

#ifdef USE_VIRTUAL_DESTRUCTOR virtual #endif ~neural_net()

Provides automatic cleanup of data.

unsigned int num_input_train_data()

Returns the number of inputs in each of the training patterns in the training_data.

FANN_EXTERNAL unsigned int FANN_API fann_num_input_train_data(
struct fann_train_data * data
)

Returns the number of inputs in each of the training patterns in the struct fann_train_data.

unsigned int num_output_train_data()

Returns the number of outputs in each of the training patterns in the struct fann_train_data.

FANN_EXTERNAL unsigned int FANN_API fann_num_output_train_data(
struct fann_train_data * data
)

Returns the number of outputs in each of the training patterns in the struct fann_train_data.

These functions allow you to interact with the FANN library from PHP.

The fixed point ANN is not as precise as a floating point ANN, furthermore it approximates the sigmoid function by a stepwise linear function.

void print_connections()

Will print the connections of the ann in a compact matrix, for easy viewing of the internals of the ann.

FANN_EXTERNAL void FANN_API fann_print_connections( struct fann * ann )

Will print the connections of the ann in a compact matrix, for easy viewing of the internals of the ann.

void print_error()

Prints the last error to stderr.

FANN_EXTERNAL void FANN_API fann_print_error( struct fann_error * errdat )

Prints the last error to stderr.

void print_parameters()

Prints all of the parameters and options of the neural network

FANN_EXTERNAL void FANN_API fann_print_parameters( struct fann * ann )

Prints all of the parameters and options of the ANN

These functions allow you to interact with the FANN library from Python.

Make sure to make and install the fann library first.

void randomize_weights( fann_type min_weight,
fann_type max_weight )

Give each connection a random weight between min_weight and max_weight

FANN_EXTERNAL void FANN_API fann_randomize_weights( struct fann * ann,
fann_type min_weight,
fann_type max_weight )

Give each connection a random weight between min_weight and max_weight

bool read_train_from_file( const std:: string & filename )

Reads a file that stores training data.

FANN_EXTERNAL struct fann_train_data *FANN_API fann_read_train_from_file(
const char * filename
)

Reads a file that stores training data.

void reset_errno()

Resets the last error number.

FANN_EXTERNAL void FANN_API fann_reset_errno( struct fann_error * errdat )

Resets the last error number.

void reset_errstr()

Resets the last error string.

FANN_EXTERNAL void FANN_API fann_reset_errstr( struct fann_error * errdat )

Resets the last error string.

void reset_MSE()

Resets the mean square error from the network.

FANN_EXTERNAL void FANN_API fann_reset_MSE( struct fann * ann )

Resets the mean square error from the network.

RPMs are a simple way to manage packages, and is used on many common Linux distributions such as Red Hat, Mandrake, and SuSE.

fann_type* run( fann_type * input )

Will run input through the neural network, returning an array of outputs, the number of which being equal to the number of neurons in the output layer.

FANN_EXTERNAL fann_type * FANN_API fann_run( struct fann * ann,
fann_type * input )

Will run input through the neural network, returning an array of outputs, the number of which being equal to the number of neurons in the output layer.

Running a fixed point ANN is done much like running an ordinary ANN.