php-ml/tests/NeuralNetwork/ActivationFunction/ThresholdedReLUTest.php

58 lines
1.4 KiB
PHP
Raw Normal View History

<?php
declare(strict_types=1);
namespace Phpml\Tests\NeuralNetwork\ActivationFunction;
use Phpml\NeuralNetwork\ActivationFunction\ThresholdedReLU;
use PHPUnit\Framework\TestCase;
class ThresholdedReLUTest extends TestCase
{
/**
* @dataProvider thresholdProvider
2018-10-28 06:44:52 +00:00
*
* @param float|int $value
*/
2018-10-28 06:44:52 +00:00
public function testThresholdedReLUActivationFunction(float $theta, float $expected, $value): void
{
$thresholdedReLU = new ThresholdedReLU($theta);
2018-10-28 06:44:52 +00:00
self::assertEquals($expected, $thresholdedReLU->compute($value));
}
public function thresholdProvider(): array
{
return [
[1.0, 0, 1.0],
[0.5, 3.75, 3.75],
[0.0, 0.5, 0.5],
[0.9, 0, 0.1],
];
}
/**
* @dataProvider thresholdDerivativeProvider
2018-10-28 06:44:52 +00:00
*
* @param float|int $value
*/
2018-10-28 06:44:52 +00:00
public function testThresholdedReLUDerivative(float $theta, float $expected, $value): void
{
$thresholdedReLU = new ThresholdedReLU($theta);
$activatedValue = $thresholdedReLU->compute($value);
2018-10-28 06:44:52 +00:00
self::assertEquals($expected, $thresholdedReLU->differentiate($value, $activatedValue));
}
public function thresholdDerivativeProvider(): array
{
return [
[0, 1, 1],
[0, 1, 0],
[0.5, 1, 1],
[0.5, 1, 1],
[0.5, 0, 0],
[2, 0, -1],
];
}
}