60 lines
860 B
Markdown
60 lines
860 B
Markdown
# Normalization
|
|
|
|
Normalization is the process of scaling individual samples to have unit norm.
|
|
|
|
## L2 norm
|
|
|
|
[http://mathworld.wolfram.com/L2-Norm.html](http://mathworld.wolfram.com/L2-Norm.html)
|
|
|
|
Example:
|
|
|
|
```
|
|
use Phpml\Preprocessing\Normalizer;
|
|
|
|
$samples = [
|
|
[1, -1, 2],
|
|
[2, 0, 0],
|
|
[0, 1, -1],
|
|
];
|
|
|
|
$normalizer = new Normalizer();
|
|
$normalizer->preprocess($samples);
|
|
|
|
/*
|
|
$samples = [
|
|
[0.4, -0.4, 0.81],
|
|
[1.0, 0.0, 0.0],
|
|
[0.0, 0.7, -0.7],
|
|
];
|
|
*/
|
|
|
|
```
|
|
|
|
## L1 norm
|
|
|
|
[http://mathworld.wolfram.com/L1-Norm.html](http://mathworld.wolfram.com/L1-Norm.html)
|
|
|
|
Example:
|
|
|
|
```
|
|
use Phpml\Preprocessing\Normalizer;
|
|
|
|
$samples = [
|
|
[1, -1, 2],
|
|
[2, 0, 0],
|
|
[0, 1, -1],
|
|
];
|
|
|
|
$normalizer = new Normalizer(Normalizer::NORM_L1);
|
|
$normalizer->preprocess($samples);
|
|
|
|
/*
|
|
$samples = [
|
|
[0.25, -0.25, 0.5],
|
|
[1.0, 0.0, 0.0],
|
|
[0.0, 0.5, -0.5],
|
|
];
|
|
*/
|
|
|
|
```
|