mirror of
https://github.com/Llewellynvdm/php-ml.git
synced 2024-12-02 01:48:22 +00:00
e83f7b95d5
- Backpropagation using the neuron activation functions derivative - instead of hardcoded sigmoid derivative - Added missing activation functions derivatives - Sigmoid forced for the output layer - Updated ThresholdedReLU default threshold to 0 (acts as a ReLU) - Unit tests for derivatives - Unit tests for classifiers using different activation functions - Added missing docs
58 lines
1.4 KiB
PHP
58 lines
1.4 KiB
PHP
<?php
|
|
|
|
declare(strict_types=1);
|
|
|
|
namespace Phpml\Tests\NeuralNetwork\ActivationFunction;
|
|
|
|
use Phpml\NeuralNetwork\ActivationFunction\Sigmoid;
|
|
use PHPUnit\Framework\TestCase;
|
|
|
|
class SigmoidTest extends TestCase
|
|
{
|
|
/**
|
|
* @dataProvider sigmoidProvider
|
|
*/
|
|
public function testSigmoidActivationFunction($beta, $expected, $value): void
|
|
{
|
|
$sigmoid = new Sigmoid($beta);
|
|
|
|
$this->assertEquals($expected, $sigmoid->compute($value), '', 0.001);
|
|
}
|
|
|
|
public function sigmoidProvider(): array
|
|
{
|
|
return [
|
|
[1.0, 1, 7.25],
|
|
[2.0, 1, 3.75],
|
|
[1.0, 0.5, 0],
|
|
[0.5, 0.5, 0],
|
|
[1.0, 0, -7.25],
|
|
[2.0, 0, -3.75],
|
|
];
|
|
}
|
|
|
|
/**
|
|
* @dataProvider sigmoidDerivativeProvider
|
|
*/
|
|
public function testSigmoidDerivative($beta, $expected, $value): void
|
|
{
|
|
$sigmoid = new Sigmoid($beta);
|
|
$activatedValue = $sigmoid->compute($value);
|
|
$this->assertEquals($expected, $sigmoid->differentiate($value, $activatedValue), '', 0.001);
|
|
}
|
|
|
|
public function sigmoidDerivativeProvider(): array
|
|
{
|
|
return [
|
|
[1.0, 0, -10],
|
|
[1, 0.006, -5],
|
|
[1.0, 0.25, 0],
|
|
[1, 0.006, 5],
|
|
[1.0, 0, 10],
|
|
[2.0, 0.25, 0],
|
|
[0.5, 0.246, 0.5],
|
|
[0.5, 0.241, 0.75],
|
|
];
|
|
}
|
|
}
|