php-ml/tests/Phpml/DimensionReduction/KernelPCATest.php

52 lines
2.1 KiB
PHP
Raw Normal View History

<?php
declare(strict_types=1);
2017-09-02 19:39:59 +00:00
namespace tests\Phpml\DimensionReduction;
use Phpml\DimensionReduction\KernelPCA;
use PHPUnit\Framework\TestCase;
class KernelPCATest extends TestCase
{
public function testKernelPCA()
{
// Acceptable error
$epsilon = 0.001;
// A simple example whose result is known beforehand
$data = [
[2,2], [1.5,1], [1.,1.5], [1.,1.],
[2.,1.],[2,2.5], [2.,3.], [1.5,3],
[1.,2.5], [1.,2.7], [1.,3.], [1,3],
[1,2], [1.5,2], [1.5,2.2], [1.3,1.7],
[1.7,1.3], [1.5,1.5], [1.5,1.6], [1.6,2],
[1.7,2.1], [1.3,1.3], [1.3,2.2], [1.4,2.4]
];
$transformed = [
[0.016485613899708], [-0.089805657741674], [-0.088695974245924], [-0.069761503810802],
[-0.068049558133392], [-0.054702087779187], [-0.063229228729333], [-0.06852813588679],
[-0.10098315410297], [-0.15617881000654], [-0.21266832077299], [-0.21266832077299],
[-0.039234518840831], [0.40858295942991], [0.40110375047242], [-0.10555116296691],
[-0.13128352866095], [-0.20865959471756], [-0.17531601535848], [0.4240660966961],
[0.36351946685163], [-0.14334173054136], [0.22454914091011], [0.15035027480881]];
$kpca = new KernelPCA(KernelPCA::KERNEL_RBF, null, 1, 15);
$reducedData = $kpca->fit($data);
// Due to the fact that the sign of values can be flipped
// during the calculation of eigenValues, we have to compare
// absolute value of the values
array_map(function ($val1, $val2) use ($epsilon) {
$this->assertEquals(abs($val1), abs($val2), '', $epsilon);
}, $transformed, $reducedData);
// Fitted KernelPCA object can also transform an arbitrary sample of the
// same dimensionality with the original dataset
$newData = [1.25, 2.25];
$newTransformed = [0.18956227539216];
$newTransformed2 = $kpca->transform($newData);
$this->assertEquals(abs($newTransformed[0]), abs($newTransformed2[0]), '', $epsilon);
}
}