Normalization

Normalization to the range 0..1

Below is a minimal example: the normalize function maps a value into the range from 0 to 1 using the minimum and maximum values.

Example of use


                
<?php

function normalize(float $value, float $min, float $max): float {
    $range = $max - $min;

    if ($range === 0.0) {
        return 0.0;
    }

    return ($value - $min) / $range;
}

try {
    $features = [1.12, 0.85, 2.33, 4.67, 2.91, 0.44, 0.58, 0.76, 3.29, 0.50];

    $min = min($features);
    $max = max($features);
    $result = [];

    foreach ($features as $feature) {
        $normalizedFeature = normalize(value: $feature, min: $min, max: $max);
        $result[] = round($normalizedFeature, 2);
    }

    echo 'Features: [1.12, 0.85, 2.33, 4.67, 2.91, 0.44, 0.58, 0.76, 3.29, 0.50]' . PHP_EOL . PHP_EOL;
    echo 'Result (normalized): ' . PHP_EOL;
    echo print_r($result, true) . PHP_EOL . PHP_EOL;
    echo 'Explanation: (1.12 - 0.44) / (4.67 - 0.44) = 0.16';
} catch (Exception $e) {
    echo 'Error: ' . $e->getMessage();
}

Result: Memory: 0.003 Mb Time running: < 0.001 sec.


                Features: [1.12, 0.85, 2.33, 4.67, 2.91, 0.44, 0.58, 0.76, 3.29, 0.50]

Result (normalized): 
Array
(
    [0] => 0.16
    [1] => 0.1
    [2] => 0.45
    [3] => 1
    [4] => 0.58
    [5] => 0
    [6] => 0.03
    [7] => 0.08
    [8] => 0.67
    [9] => 0.01
)


Explanation: (1.12 - 0.44) / (4.67 - 0.44) = 0.16

The key idea is simple: normalization makes features comparable and helps many models work more stably.