/* INTERNAL FUNCTION
Allocate room for the connections.
*/
void fann_allocate_connections(struct fann *ann)
{
ann->weights = (fann_type *) fann_calloc(ann->total_connections, sizeof(fann_type));
if(ann->weights == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return;
}
/* TODO make special cases for all places where the connections
* is used, so that it is not needed for fully connected networks.
*/
ann->connections =
(struct fann_neuron **) fann_calloc(ann->total_connections,
sizeof(struct fann_neuron *));
if(ann->connections == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return;
}
if(ann->prev_weights_deltas == NULL)
{
ann->prev_weights_deltas =
(fann_type *) fann_calloc(ann->total_connections, sizeof(fann_type));
if(ann->prev_weights_deltas == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return;
}
}
}
/* INTERNAL FUNCTION
Allocates room for the neurons.
*/
void fann_allocate_neurons(struct fann *ann)
{
struct fann_layer *layer_it;
struct fann_neuron *neurons;
unsigned int num_neurons_so_far = 0;
unsigned int num_neurons = 0;
/* all the neurons is allocated in one long array (calloc clears mem) */
neurons = (struct fann_neuron *) calloc(ann->total_neurons, sizeof(struct fann_neuron));
ann->total_neurons_allocated = ann->total_neurons;
if(neurons == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return;
}
for(layer_it = ann->first_layer; layer_it != ann->last_layer; layer_it++)
{
num_neurons = layer_it->last_neuron - layer_it->first_neuron;
layer_it->first_neuron = neurons + num_neurons_so_far;
layer_it->last_neuron = layer_it->first_neuron + num_neurons;
num_neurons_so_far += num_neurons;
}
ann->output = (fann_type *) calloc(num_neurons, sizeof(fann_type));
if(ann->output == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return;
}
}
/*
* Creates an empty set of training data
*/
FANN_EXTERNAL struct fann_train_data * FANN_API fann_create_train(unsigned int num_data, unsigned int num_input, unsigned int num_output)
{
fann_type *data_input, *data_output;
unsigned int i;
struct fann_train_data *data =
(struct fann_train_data *) malloc(sizeof(struct fann_train_data));
if(data == NULL)
{
fann_error(NULL, FANN_E_CANT_ALLOCATE_MEM);
return NULL;
}
fann_init_error_data((struct fann_error *) data);
data->num_data = num_data;
data->num_input = num_input;
data->num_output = num_output;
data->input = (fann_type **) calloc(num_data, sizeof(fann_type *));
if(data->input == NULL)
{
fann_error(NULL, FANN_E_CANT_ALLOCATE_MEM);
fann_destroy_train(data);
return NULL;
}
data->output = (fann_type **) calloc(num_data, sizeof(fann_type *));
if(data->output == NULL)
{
fann_error(NULL, FANN_E_CANT_ALLOCATE_MEM);
fann_destroy_train(data);
return NULL;
}
data_input = (fann_type *) calloc(num_input * num_data, sizeof(fann_type));
if(data_input == NULL)
{
fann_error(NULL, FANN_E_CANT_ALLOCATE_MEM);
fann_destroy_train(data);
return NULL;
}
data_output = (fann_type *) calloc(num_output * num_data, sizeof(fann_type));
if(data_output == NULL)
{
fann_error(NULL, FANN_E_CANT_ALLOCATE_MEM);
fann_destroy_train(data);
return NULL;
}
for(i = 0; i != num_data; i++)
{
data->input[i] = data_input;
data_input += num_input;
data->output[i] = data_output;
data_output += num_output;
}
return data;
}
/* INTERNAL FUNCTION
Allocates room for the neurons.
*/
void fann_allocate_neurons(struct fann *ann)
{
struct fann_layer *layer_it;
struct fann_neuron *neurons;
fann_type *sumPtr = NULL;
fann_type *valuePtr = NULL;
unsigned int num_neurons_so_far = 0;
unsigned int num_neurons = 0;
unsigned int i;
/* all the neurons is allocated in one long array (fann_calloc clears mem) */
neurons = (struct fann_neuron *) fann_calloc(ann->total_neurons, sizeof(struct fann_neuron));
if(neurons == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return;
}
// extra --- AK
sumPtr = (fann_type *)fann_calloc(ann->total_neurons, sizeof(fann_type));
valuePtr = (fann_type *)fann_calloc(ann->total_neurons, sizeof(fann_type));
for(layer_it = ann->first_layer; layer_it != ann->last_layer; layer_it++)
{
num_neurons = (unsigned int)(layer_it->last_neuron - layer_it->first_neuron);
layer_it->first_neuron = neurons + num_neurons_so_far;
layer_it->last_neuron = layer_it->first_neuron + num_neurons;
//AK
layer_it->sum = sumPtr + num_neurons_so_far;
layer_it->value = valuePtr + num_neurons_so_far;
for(i = 0; i < num_neurons; i++)
{
layer_it->first_neuron[i].sumPtr = layer_it->sum + i;
layer_it->first_neuron[i].valuePtr = layer_it->value + i;
//layer_it->first_neuron[i].activation_functionPtr = &layer_it->activation_function;
//layer_it->first_neuron[i].activation_steepnessPtr = &layer_it->activation_steepness;
}
num_neurons_so_far += num_neurons;
}
ann->output = (fann_type *) fann_calloc(num_neurons, sizeof(fann_type));
if(ann->output == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return;
}
ann->train_errors = (fann_type *) fann_calloc(ann->total_neurons, sizeof(fann_type));
if(ann->train_errors == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return;
}
}
int fann_reallocate_connections(struct fann *ann, unsigned int total_connections)
{
/* The connections are allocated, but the pointers inside are
* first moved in the end of the cascade training session.
*/
#ifdef CASCADE_DEBUG
printf("realloc from %d to %d\n", ann->total_connections_allocated, total_connections);
#endif
ann->connections =
(struct fann_neuron **) realloc(ann->connections,
total_connections * sizeof(struct fann_neuron *));
if(ann->connections == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return -1;
}
ann->weights = (fann_type *) realloc(ann->weights, total_connections * sizeof(fann_type));
if(ann->weights == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return -1;
}
fann_set_train_slopes(ann,
(fann_type *) realloc(fann_get_train_slopes(ann), total_connections * sizeof(fann_type)));
if(fann_get_train_slopes(ann) == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return -1;
}
fann_set_prev_steps(ann,
(fann_type *) realloc(fann_get_prev_steps(ann),
total_connections * sizeof(fann_type)));
if(fann_get_prev_steps(ann) == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return -1;
}
fann_set_prev_train_slopes(ann,
(fann_type *) realloc(fann_get_prev_train_slopes(ann),
total_connections * sizeof(fann_type)));
if(fann_get_prev_train_slopes(ann) == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return -1;
}
fann_set_total_connections_allocated(ann, total_connections);
return 0;
}
int fann_reallocate_neurons(struct fann *ann, unsigned int total_neurons)
{
struct fann_layer *layer_it;
struct fann_neuron *neurons;
unsigned int num_neurons = 0;
unsigned int num_neurons_so_far = 0;
neurons =
(struct fann_neuron *) realloc(ann->first_layer->first_neuron,
total_neurons * sizeof(struct fann_neuron));
fann_set_total_neurons_allocated(ann, total_neurons);
if(neurons == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return -1;
}
/* Also allocate room for more train_errors */
fann_set_train_errors(ann,
(fann_type *) realloc(fann_get_train_errors(ann),
total_neurons * sizeof(fann_type)));
if(fann_get_train_errors(ann) == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return -1;
}
if(neurons != ann->first_layer->first_neuron)
{
/* Then the memory has moved, also move the pointers */
#ifdef CASCADE_DEBUG_FULL
printf("Moving neuron pointers\n");
#endif
/* Move pointers from layers to neurons */
for(layer_it = ann->first_layer; layer_it != ann->last_layer; layer_it++)
{
num_neurons = layer_it->last_neuron - layer_it->first_neuron;
layer_it->first_neuron = neurons + num_neurons_so_far;
layer_it->last_neuron = layer_it->first_neuron + num_neurons;
num_neurons_so_far += num_neurons;
}
}
return 0;
}
FANN_EXTERNAL void FANN_API fann_set_cascade_activation_steepnesses(struct fann *ann,
fann_type *
cascade_activation_steepnesses,
unsigned int
cascade_activation_steepnesses_count)
{
if(fann_get_cascade_activation_steepnesses_count(ann) != cascade_activation_steepnesses_count)
{
fann_set_cascade_activation_steepnesses_count(ann, cascade_activation_steepnesses_count);
/* reallocate mem */
ann->cascade_params->cascade_activation_steepnesses =
(fann_type *)realloc(
ann->cascade_params->cascade_activation_steepnesses,
fann_get_cascade_activation_steepnesses_count(ann) * sizeof(fann_type));
if(ann->cascade_params->cascade_activation_steepnesses == NULL)
{
fann_error((struct fann_error*)ann, FANN_E_CANT_ALLOCATE_MEM);
return;
}
}
memmove(
ann->cascade_params->cascade_activation_steepnesses,
cascade_activation_steepnesses,
fann_get_cascade_activation_steepnesses_count(ann) * sizeof(fann_type));
}
FANN_EXTERNAL struct fann *FANN_API MAKE_NAME(create_standard)(unsigned int num_layers, ...)
{
struct fann *ann;
va_list layer_sizes;
int i;
unsigned int *layers = (unsigned int *) calloc(num_layers, sizeof(unsigned int));
if(layers == NULL)
{
fann_error(NULL, FANN_E_CANT_ALLOCATE_MEM);
return NULL;
}
va_start(layer_sizes, num_layers);
for(i = 0; i < (int) num_layers; i++)
{
layers[i] = va_arg(layer_sizes, unsigned int);
}
va_end(layer_sizes);
ann = MAKE_NAME(create_standard_array)(num_layers, layers);
free(layers);
return ann;
}
//--------------------------------------------------------------------------------------------------------
//Evaluate the ann on an array of samples
void evaluateNetwork(struct fann *ann,
const double *input,
double* output,
const unsigned int numData)
{
int i,j;
unsigned int numInputs = ann->num_input;
unsigned int numOutputs = ann->num_output;
fann_type * out = NULL;
fann_type * in = (fann_type *)MALLOC(numInputs * sizeof(fann_type));
if (in == NULL)
{
fann_error(NULL, FANN_E_CANT_ALLOCATE_MEM);
}
for(i=0;i<(int)numData;++i)
{
for(j=0;j<(int)numInputs;j++)
{
in[j] = input[(j*numData)+i];
}
out = fann_run(ann,in);
for(j=0;j<(int)numOutputs;j++)
{
output[(j*numData)+i] = out[j];
}
}
FREE(in);
}
void fann_update_candidate_weights(struct fann *ann, unsigned int num_data)
{
struct fann_neuron *first_cand = (ann->last_layer - 1)->last_neuron + 1; /* there is an empty neuron between the actual neurons and the candidate neuron */
struct fann_neuron *last_cand = first_cand + fann_get_cascade_num_candidates(ann) - 1;
switch (ann->training_algorithm)
{
case FANN_TRAIN_RPROP:
fann_update_weights_irpropm(ann, first_cand->first_con,
last_cand->last_con + ann->num_output);
break;
case FANN_TRAIN_SARPROP:
/* TODO: increase epoch? */
fann_update_weights_sarprop(ann, ann->sarprop_epoch, first_cand->first_con,
last_cand->last_con + ann->num_output);
break;
case FANN_TRAIN_QUICKPROP:
fann_update_weights_quickprop(ann, num_data, first_cand->first_con,
last_cand->last_con + ann->num_output);
break;
case FANN_TRAIN_BATCH:
case FANN_TRAIN_INCREMENTAL:
fann_error((struct fann_error *) ann, FANN_E_CANT_USE_TRAIN_ALG);
break;
}
}
/* add a layer at the position pointed to by *layer */
struct fann_layer *fann_add_layer(struct fann *ann, struct fann_layer *layer)
{
int layer_pos = (int)(layer - ann->first_layer);
int num_layers = (int)(ann->last_layer - ann->first_layer + 1);
int i;
/* allocate the layer */
struct fann_layer *layers =
(struct fann_layer *) realloc(ann->first_layer, num_layers * sizeof(struct fann_layer));
if(layers == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return NULL;
}
/* copy layers so that the free space is at the right location */
for(i = num_layers - 1; i >= layer_pos; i--)
{
layers[i] = layers[i - 1];
}
/* the newly allocated layer is empty */
layers[layer_pos].first_neuron = layers[layer_pos + 1].first_neuron;
layers[layer_pos].last_neuron = layers[layer_pos + 1].first_neuron;
/* Set the ann pointers correctly */
ann->first_layer = layers;
ann->last_layer = layers + num_layers;
#ifdef CASCADE_DEBUG_FULL
printf("add layer at pos %d\n", layer_pos);
#endif
return layers + layer_pos;
}
FANN_EXTERNAL struct fann *FANN_API fann_create_sparse(float connection_rate,
unsigned int num_layers, ...)
{
struct fann *ann;
va_list layer_sizes;
int i;
unsigned int *layers = (unsigned int *) calloc(num_layers, sizeof(unsigned int));
if(layers == NULL)
{
fann_error(NULL, FANN_E_CANT_ALLOCATE_MEM);
return NULL;
}
va_start(layer_sizes, num_layers);
for(i = 0; i < (int) num_layers; i++)
{
layers[i] = va_arg(layer_sizes, unsigned int);
}
va_end(layer_sizes);
ann = fann_create_sparse_array(connection_rate, num_layers, layers);
free(layers);
return ann;
}
/* Allocate memory for training for a layer */
FANN_EXTERNAL void FANN_API fann_layer_train_initialize_fully_recurrent(struct fann *ann, struct fann_layer *layer)
{
fann_type *free_train_errors;
struct fann_neuron *neuron_it, *last_neuron;
last_neuron = layer->last_neuron;
/* reset the layer train errors array */
if(layer->train_errors == NULL)
{
if( (layer->train_errors = (fann_type *) calloc(layer->num_outputs, sizeof(fann_type))) == NULL )
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return;
}
/* assign to MIMO neurons a piece of layer train_errors array */
free_train_errors = layer->train_errors;
for (neuron_it = layer->first_neuron; neuron_it != last_neuron; neuron_it++)
{
neuron_it->train_errors = free_train_errors;
free_train_errors += neuron_it->num_outputs;
}
}
else
{
/* clear the error variables */
memset(layer->train_errors, 0, (layer->num_outputs) * sizeof(fann_type));
}
for (neuron_it = layer->first_neuron; neuron_it != last_neuron; neuron_it++)
{
neuron_it->train_initialize(ann, layer, neuron_it);
}
}
FANN_EXTERNAL void FANN_API fann_set_cascade_activation_functions(struct fann *ann,
enum fann_activationfunc_enum *
cascade_activation_functions,
unsigned int
cascade_activation_functions_count)
{
if(fann_get_cascade_activation_functions_count(ann) != cascade_activation_functions_count)
{
fann_set_cascade_activation_functions_count(ann, cascade_activation_functions_count);
/* reallocate mem */
ann->cascade_params->cascade_activation_functions =
(enum fann_activationfunc_enum *)realloc(ann->cascade_params->cascade_activation_functions,
fann_get_cascade_activation_functions_count(ann) * sizeof(enum fann_activationfunc_enum));
if(fann_get_cascade_activation_functions(ann) == NULL)
{
fann_error((struct fann_error*)ann, FANN_E_CANT_ALLOCATE_MEM);
return;
}
}
memmove(
ann->cascade_params->cascade_activation_functions,
cascade_activation_functions,
fann_get_cascade_activation_functions_count(ann) * sizeof(enum fann_activationfunc_enum));
}
/*
* INTERNAL FUNCTION Reads training data from a file descriptor.
*/
struct fann_train_data *fann_read_train_from_fd(FILE * file, const char *filename)
{
unsigned int num_input, num_output, num_data, i, j;
unsigned int line = 1;
struct fann_train_data *data;
if(fscanf(file, "%u %u %u\n", &num_data, &num_input, &num_output) != 3)
{
fann_error(NULL, FANN_E_CANT_READ_TD, filename, line);
return NULL;
}
line++;
data = fann_create_train(num_data, num_input, num_output);
if(data == NULL)
{
return NULL;
}
for(i = 0; i != num_data; i++)
{
for(j = 0; j != num_input; j++)
{
if(!fann_scanvalue(file, FANNSCANF, &data->input[i][j]))
{
fann_error(NULL, FANN_E_CANT_READ_TD, filename, line);
fann_destroy_train(data);
return NULL;
}
}
line++;
for(j = 0; j != num_output; j++)
{
if(!fann_scanvalue(file, FANNSCANF, &data->output[i][j]))
{
fann_error(NULL, FANN_E_CANT_READ_TD, filename, line);
fann_destroy_train(data);
return NULL;
}
}
line++;
}
return data;
}
/* INTERNAL FUNCTION
compute the error at the network output
(usually, after forward propagation of a certain input vector, fann_run)
the error is a sum of squares for all the output units
also increments a counter because MSE is an average of such errors
After this train_errors in the output layer will be set to:
neuron_value_derived * (desired_output - neuron_value)
*/
void fann_compute_MSE(struct fann *ann, fann_type * desired_output)
{
fann_type neuron_value, neuron_diff, *error_it = 0, *error_begin = 0;
struct fann_neuron *last_layer_begin = (ann->last_layer - 1)->first_neuron;
const struct fann_neuron *last_layer_end = last_layer_begin + ann->num_output;
const struct fann_neuron *first_neuron = ann->first_layer->first_neuron;
/* if no room allocated for the error variabels, allocate it now */
if(ann->train_errors == NULL)
{
ann->train_errors = (fann_type *) calloc(ann->total_neurons, sizeof(fann_type));
if(ann->train_errors == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return;
}
}
else
{
/* clear the error variabels */
memset(ann->train_errors, 0, (ann->total_neurons) * sizeof(fann_type));
}
error_begin = ann->train_errors;
#ifdef DEBUGTRAIN
printf("\ncalculate errors\n");
#endif
/* calculate the error and place it in the output layer */
error_it = error_begin + (last_layer_begin - first_neuron);
for(; last_layer_begin != last_layer_end; last_layer_begin++)
{
neuron_value = last_layer_begin->value;
neuron_diff = *desired_output - neuron_value;
neuron_diff = fann_update_MSE(ann, last_layer_begin, neuron_diff);
if(ann->train_error_function)
{ /* TODO make switch when more functions */
if(neuron_diff < -.9999999)
neuron_diff = -17.0;
else if(neuron_diff > .9999999)
neuron_diff = 17.0;
else
neuron_diff = (fann_type) log((1.0 + neuron_diff) / (1.0 - neuron_diff));
}
*error_it = fann_activation_derived(last_layer_begin->activation_function,
last_layer_begin->activation_steepness, neuron_value,
last_layer_begin->sum) * neuron_diff;
desired_output++;
error_it++;
ann->num_MSE++;
}
}
int fann_check_input_output_sizes(struct fann *ann, struct fann_train_data *data)
{
if(ann->num_input != data->num_input)
{
fann_error((struct fann_error *) ann, FANN_E_INPUT_NO_MATCH,
ann->num_input, data->num_input);
return -1;
}
if(ann->num_output != data->num_output)
{
fann_error((struct fann_error *) ann, FANN_E_OUTPUT_NO_MATCH,
ann->num_output, data->num_output);
return -1;
}
return 0;
}
FANN_EXTERNAL struct fann_neuron* FANN_API fann_get_neuron_layer(struct fann *ann, struct fann_layer* layer, int neuron)
{
if(neuron >= (layer->last_neuron - layer->first_neuron))
{
fann_error((struct fann_error *) ann, FANN_E_INDEX_OUT_OF_BOUND, neuron);
return NULL;
}
return layer->first_neuron + neuron;
}
FANN_EXTERNAL struct fann_layer* FANN_API fann_get_layer(struct fann *ann, int layer)
{
if(layer <= 0 || layer >= (ann->last_layer - ann->first_layer))
{
fann_error((struct fann_error *) ann, FANN_E_INDEX_OUT_OF_BOUND, layer);
return NULL;
}
return ann->first_layer + layer;
}
/*
* Reads training data from a file.
*/
FANN_EXTERNAL struct fann_train_data *FANN_API fann_read_train_from_file(const char *configuration_file)
{
struct fann_train_data *data;
FILE *file = fopen(configuration_file, "r");
if(!file)
{
fann_error(NULL, FANN_E_CANT_OPEN_CONFIG_R, configuration_file);
return NULL;
}
data = fann_read_train_from_fd(file, configuration_file);
fclose(file);
return data;
}
/* INTERNAL FUNCTION
Save the train data structure.
*/
int fann_save_train_internal(struct fann_train_data *data, const char *filename)
{
int retval = 0;
FILE *file = fopen(filename, "w");
if(!file)
{
fann_error((struct fann_error *) data, FANN_E_CANT_OPEN_TD_W, filename);
return -1;
}
retval = fann_save_train_internal_fd(data, file, filename);
fclose(file);
return retval;
}
/*
* Scale input and output data based on previously calculated parameters.
*/
FANN_EXTERNAL void FANN_API fann_scale_train( struct fann *ann, struct fann_train_data *data )
{
unsigned cur_sample;
if(ann->scale_mean_in == NULL)
{
fann_error( (struct fann_error *) ann, FANN_E_SCALE_NOT_PRESENT );
return;
}
/* Check that we have good training data. */
/* No need for if( !params || !ann ) */
if( data->num_input != ann->num_input
|| data->num_output != ann->num_output
)
{
fann_error( (struct fann_error *) ann, FANN_E_TRAIN_DATA_MISMATCH );
return;
}
for( cur_sample = 0; cur_sample < data->num_data; cur_sample++ )
{
fann_scale_input( ann, data->input[ cur_sample ] );
fann_scale_output( ann, data->output[ cur_sample ] );
}
}
/*
* Scale input and output data based on previously calculated parameters.
*/
FANN_EXTERNAL void FANN_API fann_descale_train( struct fann *ann, struct fann_train_data *data )
{
unsigned cur_sample;
if(ann->scale_mean_in == NULL)
{
fann_error( (struct fann_error *) ann, FANN_E_SCALE_NOT_PRESENT );
return;
}
/* Check that we have good training data. */
if(fann_check_input_output_sizes(ann, data) == -1)
return;
for( cur_sample = 0; cur_sample < data->num_data; cur_sample++ )
{
fann_descale_input( ann, data->input[ cur_sample ] );
fann_descale_output( ann, data->output[ cur_sample ] );
}
}
/* INTERNAL FUNCTION
Calculates the derived of a value, given an activation function
and a steepness
*/
fann_type fann_activation_derived(unsigned int activation_function,
fann_type steepness, fann_type value, fann_type sum)
{
switch (activation_function)
{
case FANN_LINEAR:
case FANN_LINEAR_PIECE:
case FANN_LINEAR_PIECE_SYMMETRIC:
return (fann_type) fann_linear_derive(steepness, value);
case FANN_SIGMOID:
case FANN_SIGMOID_STEPWISE:
value = fann_clip(value, 0.01f, 0.99f);
return (fann_type) fann_sigmoid_derive(steepness, value);
case FANN_SIGMOID_SYMMETRIC:
case FANN_SIGMOID_SYMMETRIC_STEPWISE:
value = fann_clip(value, -0.98f, 0.98f);
return (fann_type) fann_sigmoid_symmetric_derive(steepness, value);
case FANN_GAUSSIAN:
/* value = fann_clip(value, 0.01f, 0.99f); */
return (fann_type) fann_gaussian_derive(steepness, value, sum);
case FANN_GAUSSIAN_SYMMETRIC:
/* value = fann_clip(value, -0.98f, 0.98f); */
return (fann_type) fann_gaussian_symmetric_derive(steepness, value, sum);
case FANN_ELLIOT:
value = fann_clip(value, 0.01f, 0.99f);
return (fann_type) fann_elliot_derive(steepness, value, sum);
case FANN_ELLIOT_SYMMETRIC:
value = fann_clip(value, -0.98f, 0.98f);
return (fann_type) fann_elliot_symmetric_derive(steepness, value, sum);
case FANN_SIN_SYMMETRIC:
return (fann_type) fann_sin_symmetric_derive(steepness, sum);
case FANN_COS_SYMMETRIC:
return (fann_type) fann_cos_symmetric_derive(steepness, sum);
case FANN_SIN:
return (fann_type) fann_sin_derive(steepness, sum);
case FANN_COS:
return (fann_type) fann_cos_derive(steepness, sum);
case FANN_THRESHOLD:
fann_error(NULL, FANN_E_CANT_TRAIN_ACTIVATION);
case FANN_MAXPOOLING:
return (fann_type) 1;
}
return 0;
}
/*
* Scale data in output vector before feed it to ann based on previously calculated parameters.
*/
FANN_EXTERNAL void FANN_API fann_scale_output( struct fann *ann, fann_type *output_vector )
{
unsigned cur_neuron;
if(ann->scale_mean_in == NULL)
{
fann_error( (struct fann_error *) ann, FANN_E_SCALE_NOT_PRESENT );
return;
}
for( cur_neuron = 0; cur_neuron < ann->num_output; cur_neuron++ )
output_vector[ cur_neuron ] =
(
( output_vector[ cur_neuron ] - ann->scale_mean_out[ cur_neuron ] )
/ ann->scale_deviation_out[ cur_neuron ]
- ( (fann_type)-1.0 ) /* This is old_min */
)
* ann->scale_factor_out[ cur_neuron ]
+ ann->scale_new_min_out[ cur_neuron ];
}
void fann_sparse_neuron_standard_backprop_update(struct fann *ann, struct fann_neuron *neuron)
{
unsigned int o, j;
fann_type *weights, *deltas, *mask;
const unsigned int num_outputs = neuron->num_outputs;
const unsigned int num_inputs = neuron->num_inputs;
float learning_rate = ann->backprop_params->learning_rate;
#ifdef FIXEDFANN
unsigned int decimal_point = ann->fixed_params->decimal_point;
#endif
if (neuron->num_backprop_done==0)
{
fann_error(NULL, FANN_E_CANT_USE_TRAIN_ALG);
return;
}
learning_rate=learning_rate/neuron->num_backprop_done;
/* some assignments to speed up things */
weights = neuron->weights;
deltas = neuron->weights_deltas;
mask = ((struct fann_sparse_neuron_private_data*)neuron->private_data)->mask;
for (o = 0; o < num_outputs; o++)
{
for (j = 0; j < num_inputs; j++)
{
/* adjust the weight */
weights[j] += deltas[j] * mask[j] * learning_rate; /* FIXME add the learning momentum here */
deltas[j]=0;
}
weights += num_inputs;
deltas += num_inputs;
mask += num_inputs;
}
neuron->num_backprop_done=0;
}
/* Allocate memory for training a neuron */
FANN_EXTERNAL void FANN_API fann_neuron_train_initialize_fully_recurrent(
struct fann *ann,
struct fann_layer *layer,
struct fann_neuron *neuron)
{
neuron->num_backprop_done=0;
/* allocate the weights_deltas */
if(neuron->weights_deltas == NULL)
{
if ((neuron->weights_deltas =
(fann_type*) calloc(neuron->num_weights, sizeof(fann_type))) == NULL)
{
fann_error((struct fann_error *) ann, FANN_E_CANT_ALLOCATE_MEM);
return;
}
}
else
{
/* clear the error variables */
memset(neuron->weights_deltas, 0, neuron->num_weights*sizeof(fann_type));
}
}
FANN_EXTERNAL int FANN_API fann_set_output_scaling_params(
struct fann *ann,
const struct fann_train_data *data,
float new_output_min,
float new_output_max)
{
unsigned cur_neuron, cur_sample;
/* Check that we have good training data. */
/* No need for if( !params || !ann ) */
if(data->num_input != ann->num_input
|| data->num_output != ann->num_output)
{
fann_error( (struct fann_error *) ann, FANN_E_TRAIN_DATA_MISMATCH );
return -1;
}
if(ann->scale_mean_out == NULL)
fann_allocate_scale(ann);
if(ann->scale_mean_out == NULL)
return -1;
if( !data->num_data )
{
SCALE_RESET( scale_mean, out, 0.0 )
SCALE_RESET( scale_deviation, out, 1.0 )
SCALE_RESET( scale_new_min, out, -1.0 )
SCALE_RESET( scale_factor, out, 1.0 )
}
else
{
SCALE_SET_PARAM( out );
}
return 0;
}
float fann_train_outputs_epoch(struct fann *ann, struct fann_train_data *data)
{
unsigned int i;
fann_reset_MSE(ann);
for(i = 0; i < data->num_data; i++)
{
fann_run(ann, data->input[i]);
fann_compute_MSE(ann, data->output[i]);
fann_update_slopes_batch(ann, ann->last_layer - 1, ann->last_layer - 1);
}
switch (ann->training_algorithm)
{
case FANN_TRAIN_RPROP:
fann_update_weights_irpropm(ann, (ann->last_layer - 1)->first_neuron->first_con,
ann->total_connections);
break;
case FANN_TRAIN_SARPROP:
fann_update_weights_sarprop(ann, ann->sarprop_epoch, (ann->last_layer - 1)->first_neuron->first_con,
ann->total_connections);
++(ann->sarprop_epoch);
break;
case FANN_TRAIN_QUICKPROP:
fann_update_weights_quickprop(ann, data->num_data,
(ann->last_layer - 1)->first_neuron->first_con,
ann->total_connections);
break;
case FANN_TRAIN_BATCH:
case FANN_TRAIN_INCREMENTAL:
fann_error((struct fann_error *) ann, FANN_E_CANT_USE_TRAIN_ALG);
}
return fann_get_MSE(ann);
}
FANN_EXTERNAL struct fann_train_data *FANN_API fann_subset_train_data(struct fann_train_data
*data, unsigned int pos,
unsigned int length)
{
unsigned int i;
fann_type *data_input, *data_output;
struct fann_train_data *dest =
(struct fann_train_data *) malloc(sizeof(struct fann_train_data));
if(dest == NULL)
{
fann_error((struct fann_error*)data, FANN_E_CANT_ALLOCATE_MEM);
return NULL;
}
if(pos > data->num_data || pos+length > data->num_data)
{
fann_error((struct fann_error*)data, FANN_E_TRAIN_DATA_SUBSET, pos, length, data->num_data);
return NULL;
}
fann_init_error_data((struct fann_error *) dest);
dest->error_log = data->error_log;
dest->num_data = length;
dest->num_input = data->num_input;
dest->num_output = data->num_output;
dest->input = (fann_type **) calloc(dest->num_data, sizeof(fann_type *));
if(dest->input == NULL)
{
fann_error((struct fann_error*)data, FANN_E_CANT_ALLOCATE_MEM);
fann_destroy_train(dest);
return NULL;
}
dest->output = (fann_type **) calloc(dest->num_data, sizeof(fann_type *));
if(dest->output == NULL)
{
fann_error((struct fann_error*)data, FANN_E_CANT_ALLOCATE_MEM);
fann_destroy_train(dest);
return NULL;
}
data_input = (fann_type *) calloc(dest->num_input * dest->num_data, sizeof(fann_type));
if(data_input == NULL)
{
fann_error((struct fann_error*)data, FANN_E_CANT_ALLOCATE_MEM);
fann_destroy_train(dest);
return NULL;
}
memcpy(data_input, data->input[pos], dest->num_input * dest->num_data * sizeof(fann_type));
data_output = (fann_type *) calloc(dest->num_output * dest->num_data, sizeof(fann_type));
if(data_output == NULL)
{
fann_error((struct fann_error*)data, FANN_E_CANT_ALLOCATE_MEM);
fann_destroy_train(dest);
return NULL;
}
memcpy(data_output, data->output[pos], dest->num_output * dest->num_data * sizeof(fann_type));
for(i = 0; i != dest->num_data; i++)
{
dest->input[i] = data_input;
data_input += dest->num_input;
dest->output[i] = data_output;
data_output += dest->num_output;
}
return dest;
}