wellton/php-fann
1
0
Fork 0
Código Issues Pull Requests Ações Pacotes Projetos Versões Wiki Atividade
Arquivos
master
php-fann/utils/parameters_generator.php
T
Jakub Zelenka 517e12c44c Move parameters scripts to the new directory utils
2013-04-27 18:30:46 +01:00

412 linhas
14 KiB
PHP
Original Link permanente Anotar Histórico

#!/usr/bin/env php
<?php
$training_params = array(
'training_algorithm' => array(
'params' => array('training_algorithm' => 'int'),
'comment' => 'the training algorithm as described by fann_train_enum',
'test_param' => 'FANN_TRAIN_QUICKPROP',
),
'learning_rate' => array(
'params' => array('learning_rate' => 'float'),
'comment' => 'the learning rate',
'test_param' => '0.5',
),
'learning_momentum' => array(
'params' => array('learning_momentum' => 'float'),
'comment' => 'the learning momentum',
'test_param' => '0.2',
),
'activation_function' => array(
'params' => array('activation_function' => 'int', 'layer' => 'int', 'neuron' => 'int'),
'comment' => 'the activation function for neuron number neuron in layer number layer, counting the input layer as layer 0',
'test_param' => array('FANN_SIGMOID_SYMMETRIC', '1', '1'),
'empty_function' => true,
),
'activation_function_layer' => array(
'params' => array('activation_function' => 'int', 'layer' => 'int'),
'comment' => 'the activation function for all the neurons in the layer number layer, counting the input layer as layer 0',
'only_setter' => true,
'empty_function' => true,
),
'activation_function_hidden' => array(
'params' => array('activation_function' => 'int'),
'comment' => 'the activation function for all of the hidden layers',
'only_setter' => true,
),
'activation_function_output' => array(
'params' => array('activation_function' => 'int'),
'comment' => 'the activation function for the output layer',
'only_setter' => true,
),
'activation_steepness' => array(
'params' => array('steepness' => 'fann_type', 'layer' => 'int', 'neuron' => 'int'),
'comment' => 'the activation steepness for neuron number neuron in layer number layer, counting the input layer as layer 0',
'test_param' => array('0.5', '1', '1'),
'empty_function' => true,
),
'activation_steepness_layer' => array(
'params' => array('steepness' => 'fann_type', 'layer' => 'int'),
'comment' => 'the activation steepness for all the neurons in the layer number layer, counting the input layer as layer 0',
'only_setter' => true,
'empty_function' => true,
),
'activation_steepness_hidden' => array(
'params' => array('steepness' => 'fann_type'),
'comment' => 'the activation steepness for all of the hidden layers',
'only_setter' => true,
),
'activation_steepness_output' => array(
'params' => array('steepness' => 'fann_type'),
'comment' => 'the activation steepness for the output layer',
'only_setter' => true,
),
'train_error_function' => array(
'params' => array('train_error_function' => 'int'),
'comment' => 'the error function used during training',
'test_param' => 'FANN_ERRORFUNC_LINEAR',
),
'train_stop_function' => array(
'params' => array('train_stop_function' => 'int'),
'comment' => 'the the stop function used during training',
'test_param' => 'FANN_STOPFUNC_BIT',
),
'bit_fail_limit' => array(
'params' => array('bit_fail_limit' => 'fann_type'),
'comment' => 'the bit fail limit used during training',
'test_param' => '0.5',
),
'quickprop_decay' => array(
'params' => array('quickprop_decay' => 'float'),
'comment' => 'the decay is a small negative valued number which is the factor that the weights should become smaller in each iteration during quickprop training',
'test_param' => '-0.0002',
),
'quickprop_mu' => array(
'params' => array('quickprop_mu' => 'float'),
'comment' => 'the mu factor is used to increase and decrease the step-size during quickprop training',
'test_param' => '1.7',
),
'rprop_increase_factor' => array(
'params' => array('rprop_increase_factor' => 'float'),
'comment' => 'the increase factor is a value larger than 1, which is used to increase the step-size during RPROP training',
'test_param' => '1.2',
),
'rprop_decrease_factor' => array(
'params' => array('rprop_decrease_factor' => 'float'),
'comment' => 'the decrease factor is a value smaller than 1, which is used to decrease the step-size during RPROP training',
'test_param' => '0.6',
),
'rprop_delta_min' => array(
'params' => array('rprop_delta_min' => 'float'),
'comment' => 'the minimum step-size is a small positive number determining how small the minimum step-size may be',
'test_param' => '0.1',
),
'rprop_delta_max' => array(
'params' => array('rprop_delta_max' => 'float'),
'comment' => 'the maximum step-size is a positive number determining how large the maximum step-size may be',
'test_param' => '49.5',
),
'rprop_delta_zero' => array(
'params' => array('rprop_delta_zero' => 'float'),
'comment' => 'the initial step-size is a positive number determining the initial step size',
'test_param' => '0.2',
),
'sarprop_weight_decay_shift' => array(
'params' => array('sarprop_weight_decay_shift' => 'float'),
'comment' => 'the sarprop weight decay shift',
'test_param' => '-6.744',
),
'sarprop_step_error_threshold_factor' => array(
'params' => array('sarprop_step_error_threshold_factor' => 'float'),
'comment' => 'the sarprop step error threshold factor',
'test_param' => '0.2',
),
'sarprop_step_error_shift' => array(
'params' => array('sarprop_step_error_shift' => 'float'),
'comment' => 'the get sarprop step error shift',
'test_param' => '1.4',
),
'sarprop_temperature' => array(
'params' => array('sarprop_temperature' => 'float'),
'comment' => 'the sarprop weight decay shift',
'test_param' => '0.02',
),
);
$cascade_params = array(
'cascade_output_change_fraction' => array(
'params' => array( 'cascade_output_change_fraction' => 'float' ),
'comment' => 'the cascade output change fraction',
'test_param' => 0.02,
),
'cascade_output_stagnation_epochs' => array(
'params' => array( 'cascade_output_stagnation_epochs' => 'int' ),
'comment' => 'the number of cascade output stagnation epochs',
'test_param' => 10,
),
'cascade_candidate_change_fraction' => array(
'params' => array( 'cascade_candidate_change_fraction' => 'float' ),
'comment' => 'the cascade candidate change fraction',
'test_param' => 0.02,
),
'cascade_candidate_stagnation_epochs' => array(
'params' => array( 'cascade_candidate_stagnation_epochs' => 'int' ),
'comment' => 'the number of cascade candidate stagnation epochs',
'test_param' => 10,
),
'cascade_weight_multiplier' => array(
'params' => array( 'cascade_weight_multiplier' => 'fann_type' ),
'comment' => 'the weight multiplier',
'test_param' => 0.5,
),
'cascade_candidate_limit' => array(
'params' => array( 'cascade_candidate_limit' => 'fann_type' ),
'comment' => 'the candidate limit',
'test_param' => 950.0,
),
'cascade_max_out_epochs' => array(
'params' => array( 'cascade_max_out_epochs' => 'int' ),
'comment' => 'the maximum out epochs',
'test_param' => 140,
),
'cascade_min_out_epochs' => array(
'params' => array( 'cascade_min_out_epochs' => 'int' ),
'comment' => 'the minimum out epochs',
'test_param' => 60,
),
'cascade_max_cand_epochs' => array(
'params' => array( 'cascade_max_cand_epochs' => 'int' ),
'comment' => 'the max candidate epochs',
'test_param' => 140,
),
'cascade_min_cand_epochs' => array(
'params' => array( 'cascade_min_cand_epochs' => 'int' ),
'comment' => 'the min candidate epochs',
'test_param' => 60,
),
'cascade_num_candidates' => array(
'params' => array( 'cascade_num_candidates' => 'int' ),
'comment' => 'the number of candidates used during training',
'only_getter' => true
),
'cascade_activation_functions_count' => array(
'params' => array( 'cascade_activation_functions_count' => 'int' ),
'comment' => 'the number of activation functions',
'only_getter' => true
),
'cascade_activation_functions' => array(
'params' => array( 'cascade_activation_functions' => 'array' ),
'comment' => 'the array of cascade candidate activation functions',
'empty_function' => true
),
'cascade_activation_steepnesses_count' => array(
'params' => array( 'cascade_activation_steepnesses_count' => 'int' ),
'comment' => 'the number of activation steepnesses',
'only_getter' => true
),
'cascade_activation_steepnesses' => array(
'params' => array( 'cascade_activation_steepnesses' => 'array' ),
'comment' => 'the cascade activation steepnesses array is an array of the different activation functions used by the candidates',
'empty_function' => true
),
'cascade_num_candidate_groups' => array(
'params' => array( 'cascade_num_candidate_groups' => 'int' ),
'comment' => 'the number of candidate groups',
'test_param' => 3,
),
);
$fpg_stream = STDOUT;
function fpg_printf($format = "") {
global $fpg_stream;
$args = func_get_args();
vfprintf($fpg_stream, $format . PHP_EOL, array_splice($args, 1));
}
/* print function headers */
function fpg_headers($name, $params) {
if ($params['has_getter']) {
fpg_printf('PHP_FUNCTION(fann_get_%s);', $name);
}
if ($params['has_setter']) {
fpg_printf('PHP_FUNCTION(fann_set_%s);', $name);
}
}
/* print function params definitions for reflection */
function fpg_reflections($name, $params) {
$begin = "ZEND_BEGIN_ARG_INFO(arginfo_fann_%s_$name, 0)";
$arg = "ZEND_ARG_INFO(0, %s)";
$end = "ZEND_END_ARG_INFO()";
// getter
if ($params['has_getter']) {
fpg_printf($begin, 'get');
fpg_printf($arg, 'ann');
$first = true;
foreach ($params['params'] as $param_name => $param_type) {
if ($first)
$first = false;
else
fpg_printf($arg, $param_name);
}
fpg_printf($end);
fpg_printf(); // EOL
}
// setter
if ($params['has_setter']) {
fpg_printf($begin, 'set');
fpg_printf($arg, 'ann');
foreach ($params['params'] as $param_name => $param_type) {
fpg_printf($arg, $param_name);
}
fpg_printf($end);
fpg_printf();
}
}
/* print function definitions for the module spec */
function fpg_definitions($name, $params) {
$format = "\tPHP_FE(fann_%-41s arginfo_fann_%s_$name)";
if ($params['has_getter'])
fpg_printf($format, "get_$name,", "get");
if ($params['has_setter'])
fpg_printf($format, "set_$name,", "set");
}
/* print functions body */
function fpg_functions($name, $params) {
$return_type = $params['return'];
$return_cmd = 'RETURN_';
switch ($return_type) {
case 'float':
case 'double':
case 'fann_type':
$return_cmd .= 'DOUBLE';
break;
case 'int':
$return_cmd .= 'LONG';
break;
}
if ($params['has_getter']) {
$body = $params['empty_function'] ? "": "PHP_FANN_GET_PARAM(fann_get_$name, $return_cmd);";
$getter = "
/* {{{ proto bool fann_get_$name(resource ann)
Returns {$params['comment']} */
PHP_FUNCTION(fann_get_$name)
{
$body
}
/* }}} */";
fpg_printf($getter);
}
if ($params['has_setter']) {
$set_args = "fann_set_$name";
$comment_args = "resource ann";
foreach ($params['params'] as $param_name => $param_type) {
$comment_args .= sprintf(", %s %s", $param_type == 'fann_type' ? 'double' : $param_type, $param_name);
$set_args .= sprintf(", %s", $param_type == 'int' ? 'l, long' : 'd, double');
}
$body = $params['empty_function'] ? "": "PHP_FANN_SET_PARAM($set_args);";
$setter = "
/* {{{ proto bool fann_set_$name($comment_args)
Sets {$params['comment']}*/
PHP_FUNCTION(fann_set_$name)
{
$body
}
/* }}} */";
fpg_printf($setter);
}
}
/* generate testes */
function fpg_tests($name, $params) {
$file_path = (dirname(__FILE__) . "/tests/fann_%s_{$name}_basic.phpt");
if (is_file($file_path) && $params['empty_test'])
return;
$test_params = isset($params['test_param']) ? $params['test_param'] : '';
if (is_array($test_params)) {
$setter_test_param = implode(', ', $test_params);
$setter_first_param = array_shift($test_params);
$getter_test_param = ', ' . implode(', ', $test_params);
} else {
$setter_test_param = $setter_first_param = $test_params;
$getter_test_param = '';
}
// getter
$getter_test_file = sprintf($file_path, 'get');
if ((true || !is_file($file_path) || !$params['empty_test']) && $params['has_getter']) {
$getter_test_content = '--TEST--
Test function fann_get_' . $name . '() by calling it with its expected arguments
--FILE--
<?php
$ann = fann_create_standard( 3, 3, 2, 1 );
var_dump( fann_get_' . $name . '( $ann' . $getter_test_param . ' ) );
?>
--EXPECTF--
' . ($params['return'] == 'int' ? 'int(%d)' : ($params['return'] == 'array' ? 'array(%d) {%A}' : 'float(%f)'));
if ($params['empty_test']) {
$getter_test_content = (substr($getter_test_content, 0, strpos($getter_test_content, '<?php') + 6) .
substr($getter_test_content, strpos($getter_test_content, '?>') - 1));
}
file_put_contents($getter_test_file, $getter_test_content);
}
// setter
$setter_test_file = sprintf($file_path, 'set');
if ((true || !is_file($file_path) || !$params['empty_test']) && $params['has_setter'] && $setter_test_param) {
if (in_array($params['return'], array('float', 'double', 'fann_type')))
$setter_cmp_cond = 'round( fann_get_' .$name . '( $ann' . $getter_test_param . ' ), 2 ) == round( ' . $setter_first_param . ', 2 )';
else
$setter_cmp_cond = 'fann_get_' . $name . '( $ann' . $getter_test_param . ' ) == ' . $setter_first_param;
$setter_test_content = '--TEST--
Test function fann_set_' . $name . '() by calling it with its expected arguments
--FILE--
<?php
$ann = fann_create_standard( 3, 3, 2, 1 );
var_dump( fann_set_' . $name . '( $ann, ' . $setter_test_param . ' ) );
var_dump( ' . $setter_cmp_cond . ' );
?>
--EXPECT--
bool(true)
bool(true)';
file_put_contents($setter_test_file, $setter_test_content);
}
}
/* process parameters array */
function fpg_process_array($action, $params_list) {
$fce_name = "fpg_$action";
if (function_exists($fce_name)) {
foreach ($params_list as $name => $params) {
if (!isset($params['return']))
$params['return'] = reset($params['params']);
$params['empty_function'] = isset($params['empty_function']) && $params['empty_function'];
$params['empty_test'] = isset($params['empty_test']) && $params['empty_test'];
$params['has_setter'] = !isset($params['only_getter']) || !$params['only_getter'];
$params['has_getter'] = !isset($params['only_setter']) || !$params['only_setter'];
$fce_name($name, $params);
}
}
else {
fprintf(STDERR, "Argument $action is not a valid action\n");
}
}
if ($argc < 2)
die("You need to provide arguments\n");
for ($i = 1; $i < $argc; $i++) {
fpg_process_array($argv[$i], $cascade_params);
fpg_process_array($argv[$i], $training_params);
}
Referência em uma Nova Issue Ver Git Blame Copiar Link Permanente