sgpp::datadriven::BOConfig Class Reference

Container class to store a conrete hyperparameter configuration for interaction with Bayesian Optimization. More...

#include <BOConfig.hpp>

Public Member Functions
	BOConfig ()=default
	Default Constructor. More...
	BOConfig (std::vector< int > *discOptions, std::vector< int > *catOptions, size_t nCont)
	Constructor for making a prototype based on the number of hyperparameters. More...
void	calcDiscDistance (BOConfig &other, base::DataVector &scales)
	calculation of discrete part of the distance between two BOConfigs/sample points More...
int	getCat (size_t idx)
	Get the value of a specific categorical parameter. More...
double	getCont (size_t idx)
	Get the value of a specific continuous parameter. More...
size_t	getContSize ()
	Get number of continuous parameters. More...
int	getDisc (size_t idx)
	Get the value of a specific discrete parameter. More...
size_t	getNPar () const
	Get number total number of parameters. More...
double	getScaledDistance (BOConfig &other, const base::DataVector &scales)
	Compute complete distance to another BOConfig/sample point. More...
double	getScore ()
	Get score measured on this sample. More...
double	getTotalDistance (const base::DataVector &input, base::DataVector &scales)
	finish previous distance calculation by adding the continuous part More...
bool	nextDisc ()
	Iterator over discrete parameter options. More...
void	randomize (std::mt19937 &generator)
	Generate a random config. More...
void	setCont (const base::DataVector &input)
	Set the continuous parameters according to input. More...
void	setScore (double input)
	Set score measured on this sample. More...

Detailed Description

Container class to store a conrete hyperparameter configuration for interaction with Bayesian Optimization.

Constructor & Destructor Documentation

BOConfig() [1/2]

sgpp::datadriven::BOConfig::BOConfig ( )

default

Default Constructor.

BOConfig() [2/2]

sgpp::datadriven::BOConfig::BOConfig	(	std::vector< int > *	discOptions,
		std::vector< int > *	catOptions,
		size_t	nCont
	)

Constructor for making a prototype based on the number of hyperparameters.

Parameters

discOptions	number of options for each discrete parameter
catOptions	number of options for each categorical parameter
nCont	number of continuous parameters

Member Function Documentation

calcDiscDistance()

void sgpp::datadriven::BOConfig::calcDiscDistance	(	BOConfig &	other,
		base::DataVector &	scales
	)

calculation of discrete part of the distance between two BOConfigs/sample points

Parameters

other	sample point to calculate distance to
scales	scaling of hyperparameters in relation to each other

References python.statsfileInfo::i, and sgpp::combigrid::pow().

Referenced by sgpp::datadriven::BayesianOptimization::main().

getCat()

int sgpp::datadriven::BOConfig::getCat

(

size_t

idx

)

Get the value of a specific categorical parameter.

Parameters

idx	parameter position

Returns

parameter value

Referenced by sgpp::datadriven::FitterFactory::setBO().

getCont()

double sgpp::datadriven::BOConfig::getCont

(

size_t

idx

)

Get the value of a specific continuous parameter.

Parameters

idx	parameter position

Returns

parameter value

Referenced by sgpp::datadriven::FitterFactory::setBO().

getContSize()

size_t sgpp::datadriven::BOConfig::getContSize

(

)

Get number of continuous parameters.

Returns

number of continuous parameters

Referenced by sgpp::datadriven::BayesianOptimization::main().

getDisc()

int sgpp::datadriven::BOConfig::getDisc

(

size_t

idx

)

Get the value of a specific discrete parameter.

Parameters

idx	parameter position

Returns

parameter value

Referenced by sgpp::datadriven::FitterFactory::setBO().

getNPar()

size_t sgpp::datadriven::BOConfig::getNPar

(

)

const

Get number total number of parameters.

Returns

number of continuous parameters

getScaledDistance()

double sgpp::datadriven::BOConfig::getScaledDistance	(	BOConfig &	other,
		const base::DataVector &	scales
	)

Compute complete distance to another BOConfig/sample point.

Parameters

other	sample point to calculate distance to
scales	scaling of hyperparameters in relation to each other

Returns

distance measure

References python.statsfileInfo::i, sgpp::combigrid::pow(), and analyse_erg::tmp.

getScore()

double sgpp::datadriven::BOConfig::getScore

(

)

Get score measured on this sample.

Returns

score

Referenced by sgpp::datadriven::BayesianOptimization::updateGP().

getTotalDistance()

double sgpp::datadriven::BOConfig::getTotalDistance	(	const base::DataVector &	input,
		base::DataVector &	scales
	)

finish previous distance calculation by adding the continuous part

Parameters

input	continuous part of the other (new) sample point
scales	scaling of hyperparameters in relation to each other

Returns

complete distance measure

References python.statsfileInfo::i, sgpp::combigrid::pow(), and analyse_erg::tmp.

nextDisc()

bool sgpp::datadriven::BOConfig::nextDisc

(

)

Iterator over discrete parameter options.

Returns

stopping criterion

References python.statsfileInfo::i.

Referenced by sgpp::datadriven::BayesianOptimization::main().

randomize()

void sgpp::datadriven::BOConfig::randomize

(

std::mt19937 &

generator

)

Generate a random config.

Parameters

generator

for seeded rng

References python.statsfileInfo::i.

setCont()

void sgpp::datadriven::BOConfig::setCont

(

const base::DataVector &

input

)

Set the continuous parameters according to input.

Parameters

input

DataVector holding continuous parameters

Referenced by sgpp::datadriven::BayesianOptimization::main().

setScore()

void sgpp::datadriven::BOConfig::setScore

(

double

input

)

Set score measured on this sample.

Parameters

input

score

Referenced by sgpp::datadriven::BoHyperparameterOptimizer::run().

---

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

• datadriven/src/sgpp/datadriven/datamining/modules/hpo/bo/BOConfig.hpp
• datadriven/src/sgpp/datadriven/datamining/modules/hpo/bo/BOConfig.cpp