This is a PHP wrapper for FANN (Fast Artificial Neural Network) library.
The API is documented on http://www.php.net/manual/en/book.fann.php where is the complete documentation for PHP FANN.
The API is very similar to the official FANN C API. Just functions for fixed fann_type
have not been mapped because PHP always support float
. In addition unnecessary arguments for some functions have been left out (for example array length that is not necessary for PHP arrays).
The extension can be installed on Linux and Windows.
Before you start installation make sure that libfann
is installed on your system. It's part of the main repository in the most Linux distributions (search for fann
). If not you need to install it first. Either download it from the official site or get it from your distro repository. For example on Ubuntu:
$ sudo apt-get install libfann-dev
Fann installation can be skipped if an RPM for Fedora is used (libfann
is in the package dependencies).
The RPM package for PHP FANN is available in Remi's repository: http://rpms.famillecollet.com/
It is available for Fedora, RHEL and clones (CentOS, SC and others).
After downloading remi-release RPM, the package can be installed by executing following command:
$ sudo yum --enablerepo=remi install php-pecl-fann
This extension is available on PECL. The installation is very simple. Just run:
$ sudo pecl install fann
It's important to have a git installed as it's necessary for recursive fetch of phpc.
First clone recursively the repository
git clone --recursive https://github.com/bukka/php-fann.git
Then go to the created source directory and compile the extension. You need to have a php development package installed (command phpize
must be available).
cd php-fann
phpize
./configure --with-fann
make
sudo make install
Finally you need to add
extension=fann.so
to the php.ini
Precompiled binary dll
libraries for php-fann and libfann are available on the PECL fann page. The compiled version of libfann is 2.2.
These are just two basic examples for simple training and running supplied data on the trained network.
$num_input = 2;
$num_output = 1;
$num_layers = 3;
$num_neurons_hidden = 3;
$desired_error = 0.001;
$max_epochs = 500000;
$epochs_between_reports = 1000;
$ann = fann_create_standard($num_layers, $num_input, $num_neurons_hidden, $num_output);
if ($ann) {
fann_set_activation_function_hidden($ann, FANN_SIGMOID_SYMMETRIC);
fann_set_activation_function_output($ann, FANN_SIGMOID_SYMMETRIC);
$filename = dirname(__FILE__) . "/xor.data";
if (fann_train_on_file($ann, $filename, $max_epochs, $epochs_between_reports, $desired_error))
fann_save($ann, dirname(__FILE__) . "/xor_float.net");
fann_destroy($ann);
}
$train_file = (dirname(__FILE__) . "/xor_float.net");
if (!is_file($train_file))
die("The file xor_float.net has not been created! Please run simple_train.php to generate it");
$ann = fann_create_from_file($train_file);
if (!$ann)
die("ANN could not be created");
$input = array(-1, 1);
$calc_out = fann_run($ann, $input);
printf("xor test (%f,%f) -> %f\n", $input[0], $input[1], $calc_out[0]);
fann_destroy($ann);