Phpml\NeuralNetwork\ActivationFunction\PReLU

Complexity

Total Complexity

5

Size/Duplication

Total Lines	29
Duplicated Lines	0 %

Coupling/Cohesion

Components	1
Dependencies	0

Importance

Changes

0

3 Methods

Rating	Name	Duplication	Size	Complexity
A	__construct()	0	4	1
A	compute()	0	4	2
A	differentiate()	0	4	2

<?php

declare(strict_types=1);

namespace Phpml\NeuralNetwork\ActivationFunction;

use Phpml\NeuralNetwork\ActivationFunction;

class PReLU implements ActivationFunction
{
    /**
     * @var float
     */
    private $beta;

    public function __construct(float $beta = 0.01)
    {
        $this->beta = $beta;
    }

    /**
     * @param float|int $value
     */
    public function compute($value): float
    {
        return $value >= 0 ? $value : $this->beta * $value;
    }

    /**
     * @param float|int $value
     * @param float|int $computedvalue
     */
    public function differentiate($value, $computedvalue): float
    {
        return $computedvalue >= 0 ? 1.0 : $this->beta;
    }
}