/**
* Validate a network for training.
*
* @param ContainsFlat $network
* The network to validate.
* @param MLDataSet $training
* The training set to validate.
*/
function validateNetworkForTraining(ContainsFlat $network, MLDataSet $training)
{
$inputCount = $network->getFlat()->getInputCount();
$outputCount = $network->getFlat()->getOutputCount();
if ($inputCount != $training->getInputSize()) {
throw new NeuralNetworkError("The input layer size of " + $inputCount + " must match the training input size of " + $training->getInputSize() + ".");
}
if ($training->getIdealSize() > 0 && $outputCount != $training->getIdealSize()) {
throw new NeuralNetworkError("The output layer size of " + $outputCount + " must match the training input size of " + $training->getIdealSize() + ".");
}
}
/**
* Copy whatever dataset type is specified into a memory dataset.
*
* @param
* MLDataSet set
* The dataset to copy.
*/
public function copy(MLDataSet $set)
{
$inputCount = $set->getInputSize();
$idealCount = $set->getIdealSize();
foreach ($set as $pair) {
$input = null;
$ideal = null;
if ($inputCount > 0) {
$input = new BasicMLData($inputCount);
EngineArray\arrayCopy($pair->getInputArray(), $input->getData());
}
if ($idealCount > 0) {
$ideal = new BasicMLData($idealCount);
EngineArray\arrayCopy($pair->getIdealArray(), $ideal->getData());
}
$this->add(new BasicMLDataPair($input, $ideal));
}
}
/**
* Calculate the error for this neural network.
* The error is calculated
* using root-mean-square(RMS).
*
* @param
* MLDataSet data
* The training set.
* @return double The error percentage.
*/
public function calculateError(MLDataSet $data)
{
$errorCalculation = new ErrorCalculation();
$actual = array();
$pair = BasicMLDataPair::createPair($data->getInputSize(), $data->getIdealSize());
for ($i = 0; $i < $data->getRecordCount(); ++$i) {
$data->getRecord($i, $pair);
$this->compute($pair->getInputArray(), $actual);
$errorCalculation->updateError($actual, $pair->getIdealArray(), $pair->getSignificance());
}
return $errorCalculation->calculate();
}