#include <Scaler.h>

Inheritance diagram for StandardScaler:

Public Member Functions
	StandardScaler (bool withMeans=true, bool withStd=true)

void	fit (const Matrix< double > &X, const Matrix< double > &y) override

Matrix< double >	transform (const Matrix< double > &in) override

Public Member Functions inherited from Transformer
Matrix< double >	predict (const Matrix< double > &in) override

virtual void	fit (const Matrix< double > &X, const Matrix< double > &y)=0

virtual Matrix< double >	predict (const Matrix< double > &)=0

virtual Matrix< double >	transform (const Matrix< double > &)=0

Public Attributes
Matrix< double >	means

Matrix< double >	std_deviations

Private Attributes
bool	with_std = true

bool	with_means = true

Detailed Description

Standarize features by removing the mean and scaling to unit variance.

The standard score of a sample $x\inX$ with $X\in\mathbf{R}^{N\times M}$ is calculated as $$ \tilde{x} = \frac{x - \mu}{\sigma} $$ with

$\mu\in\mathbf{R}^M$ the mean of the training samples
$\sigma\in\mathbf{R}^M$ the standard deviation of the training samples

Note: Standardization of a data set is a common requirement for many ML estimators: They might behave badly if the individual featrues do not more or less look like standard normally distributed data.

For instance many elements used in the objective funtion of a learning algorithm assume that all features are centered around 0 and have variance in the same order. If a feature has a variance that is orders of maginiteds larger than others, it might dominate the objective function and make the estimator unable to learn from other feaztures correctly as expected.[^1]

[^1]: https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html

Examples: ds/preprocessing/TestScaler.cpp.

Constructor & Destructor Documentation

◆ StandardScaler()

StandardScaler::StandardScaler	(	bool	withMeans = `true`,
		bool	withStd = `true`
	)

inline

Member Function Documentation

◆ fit()

void StandardScaler::fit	(	const Matrix< double > &	X,
		const Matrix< double > &	y
	)

inlineoverridevirtual

Computes mean and sds to be used later in scaling.

Parameters

X	given matrix to use to calculate mean/standard deviations
y	unused

Implements Predictor.

◆ transform()

Matrix< double > StandardScaler::transform ( const Matrix< double > & in )

inlineoverridevirtual

Perform standarization by centering and scaling

To achieve center at 0 and unit variance.

Parameters

in	Data to transform

Returns: : transformed data

Implements Predictor.

Member Data Documentation

◆ means

Matrix<double> StandardScaler::means

◆ std_deviations

Matrix<double> StandardScaler::std_deviations

◆ with_means

bool StandardScaler::with_means = true

private

◆ with_std

bool StandardScaler::with_std = true