StochasticTools::Standardizer Class Reference

Class for standardizing data (centering and scaling) More...

#include <Standardizer.h>

Public Member Functions
	Standardizer ()=default
void	set (const Real &n)
	Methods for setting mean and standard deviation directly Sets mean=0, std=1 for n variables. More...
void	set (const Real &mean, const Real &stdev)
	Sets mean and std for a single variable. More...
void	set (const Real &mean, const Real &stdev, const Real &n)
	Sets mean and std for a n variables variable. More...
void	set (const std::vector< Real > &mean, const std::vector< Real > &stdev)
	Sets mean and std directly using provided vectors. More...
const std::vector< Real > &	getMean () const
	Get the mean vector. More...
const std::vector< Real > &	getStdDev () const
	Get the standard deviation vector. More...
void	computeSet (const RealEigenMatrix &input)
	Methods for computing and setting mean and standard deviation. More...
void	storeHelper (std::ostream &stream, void *context) const
	Helper for dataStore. More...
void	getStandardized (RealEigenMatrix &input) const
	Returns the standardized (centered and scaled) of the provided input. More...
void	getDestandardized (RealEigenMatrix &input) const
	De-standardizes (de-centered and de-scaled) the assumed standardized input. More...
void	getDescaled (RealEigenMatrix &input) const
	De-scales the assumed scaled input. More...

Protected Attributes
std::vector< Real >	_mean
std::vector< Real >	_stdev

Detailed Description

Class for standardizing data (centering and scaling)

Definition at line 21 of file Standardizer.h.

Constructor & Destructor Documentation

Standardizer()

StochasticTools::Standardizer::Standardizer ( )

default

Member Function Documentation

computeSet()

void StochasticTools::Standardizer::computeSet

(

const RealEigenMatrix &

input

)

Methods for computing and setting mean and standard deviation.

Definition at line 58 of file Standardizer.C.

Referenced by StochasticTools::GaussianProcess::standardizeData(), and StochasticTools::GaussianProcess::standardizeParameters().

{
  _mean.clear();
  _stdev.clear();
  unsigned int num_samples = input.rows();
  unsigned int n = input.cols();
  // comptue mean
  RealEigenVector mean = input.colwise().mean();
  // Compute standard deviation
  RealEigenVector stdev =
      ((input.rowwise() - mean.transpose()).colwise().squaredNorm() / num_samples)
          .transpose()
          .array()
          .sqrt();
  // Store in std:vector format
  _mean.resize(n);
  _stdev.resize(n);
  RealEigenVector::Map(&_mean[0], n) = mean;
  RealEigenVector::Map(&_stdev[0], n) = stdev;
}

getDescaled()

void StochasticTools::Standardizer::getDescaled

(

RealEigenMatrix &

input

)

const

De-scales the assumed scaled input.

Definition at line 97 of file Standardizer.C.

Referenced by GaussianProcessSurrogate::evaluate().

{
  Eigen::Map<const RealEigenVector> stdev(_stdev.data(), _stdev.size());
  input = input.array().rowwise() * stdev.transpose().array();
}

getDestandardized()

void StochasticTools::Standardizer::getDestandardized

(

RealEigenMatrix &

input

)

const

De-standardizes (de-centered and de-scaled) the assumed standardized input.

Definition at line 88 of file Standardizer.C.

Referenced by GaussianProcessSurrogate::evaluate().

{
  Eigen::Map<const RealEigenVector> mean(_mean.data(), _mean.size());
  Eigen::Map<const RealEigenVector> stdev(_stdev.data(), _stdev.size());
  input =
      (input.array().rowwise() * stdev.transpose().array()).rowwise() + mean.transpose().array();
}

getMean()

const std::vector<Real>& StochasticTools::Standardizer::getMean ( ) const

inline

Get the mean vector.

Definition at line 37 of file Standardizer.h.

Referenced by LibtorchANNSurrogate::evaluate().

{ return _mean; }

getStandardized()

void StochasticTools::Standardizer::getStandardized

(

RealEigenMatrix &

input

)

const

Returns the standardized (centered and scaled) of the provided input.

Definition at line 80 of file Standardizer.C.

Referenced by GaussianProcessSurrogate::evaluate(), StochasticTools::GaussianProcess::standardizeData(), and StochasticTools::GaussianProcess::standardizeParameters().

{
  Eigen::Map<const RealEigenVector> mean(_mean.data(), _mean.size());
  Eigen::Map<const RealEigenVector> stdev(_stdev.data(), _stdev.size());
  input = (input.rowwise() - mean.transpose()).array().rowwise() / stdev.transpose().array();
}

getStdDev()

const std::vector<Real>& StochasticTools::Standardizer::getStdDev ( ) const

inline

Get the standard deviation vector.

Definition at line 39 of file Standardizer.h.

Referenced by LibtorchANNSurrogate::evaluate().

{ return _stdev; }

set() [1/4]

void StochasticTools::Standardizer::set

(

const Real &

n

)

Methods for setting mean and standard deviation directly Sets mean=0, std=1 for n variables.

Definition at line 16 of file Standardizer.C.

Referenced by dataLoad(), GaussianProcessTrainer::postTrain(), LibtorchANNTrainer::postTrain(), and ActiveLearningGaussianProcess::reTrain().

{
  _mean.clear();
  _stdev.clear();
  for (unsigned int ii = 0; ii < n; ++ii)
  {
    _mean.push_back(0);
    _stdev.push_back(1);
  }
}

set() [2/4]

void StochasticTools::Standardizer::set	(	const Real &	mean,
		const Real &	stdev
	)

Sets mean and std for a single variable.

Definition at line 28 of file Standardizer.C.

{
  _mean.clear();
  _stdev.clear();
  _mean.push_back(mean);
  _stdev.push_back(stdev);
}

set() [3/4]

void StochasticTools::Standardizer::set	(	const Real &	mean,
		const Real &	stdev,
		const Real &	n
	)

Sets mean and std for a n variables variable.

Definition at line 37 of file Standardizer.C.

{
  _mean.clear();
  _stdev.clear();
  for (unsigned int ii = 0; ii < n; ++ii)
  {
    _mean.push_back(mean);
    _stdev.push_back(stdev);
  }
}

set() [4/4]

void StochasticTools::Standardizer::set	(	const std::vector< Real > &	mean,
		const std::vector< Real > &	stdev
	)

Sets mean and std directly using provided vectors.

Definition at line 49 of file Standardizer.C.

{
  mooseAssert(mean.size() == stdev.size(),
              "Provided mean and standard deviation vectors are of differing size.");
  _mean = mean;
  _stdev = stdev;
}

storeHelper()

void StochasticTools::Standardizer::storeHelper	(	std::ostream &	stream,
		void *	context
	)		const

Helper for dataStore.

Definition at line 105 of file Standardizer.C.

Referenced by dataStore().

{
  unsigned int n = _mean.size();
  dataStore(stream, n, context);
  for (unsigned int ii = 0; ii < n; ++ii)
    dataStore(stream, _mean[ii], context);
  for (unsigned int ii = 0; ii < n; ++ii)
    dataStore(stream, _stdev[ii], context);
}

Member Data Documentation

_mean

std::vector<Real> StochasticTools::Standardizer::_mean

protected

Definition at line 57 of file Standardizer.h.

Referenced by computeSet(), getDestandardized(), getMean(), getStandardized(), set(), and storeHelper().

_stdev

std::vector<Real> StochasticTools::Standardizer::_stdev

protected

Definition at line 58 of file Standardizer.h.

Referenced by computeSet(), getDescaled(), getDestandardized(), getStandardized(), getStdDev(), set(), and storeHelper().

---

The documentation for this class was generated from the following files:

• stochastic_tools/include/utils/Standardizer.h
• stochastic_tools/src/utils/Standardizer.C