GP_Hedge model as describen in Hoffman et al. More...

#include <criteria_hedge.hpp>

Inheritance diagram for bayesopt::GP_Hedge:

Collaboration diagram for bayesopt::GP_Hedge:

Public Member Functions
void	init (NonParametricProcess *proc)

double	operator() (const vectord &x)

bool	requireComparison ()

void	initialCriteria ()

bool	rotateCriteria ()

void	pushResult (const vectord &prevResult)

std::string	getBestCriteria (vectord &best)

std::string	name ()

Public Member Functions inherited from bayesopt::CombinedCriteria
void	pushCriteria (Criteria *crit)

void	setParameters (const vectord &theta)

size_t	nParameters ()

Public Member Functions inherited from bayesopt::Criteria
double	evaluate (const vectord &x)

virtual void	reset ()

void	setRandomEngine (randEngine &eng)

virtual void	update (const vectord &x)

Protected Member Functions
int	update_hedge ()

virtual double	computeLoss (const vectord &query)

Protected Attributes
vectord	loss_

vectord	gain_

vectord	prob_

vectord	cumprob_

Criteria *	mCurrentCriterium

std::vector< vectord >	mBestLists

size_t	mIndex

Protected Attributes inherited from bayesopt::CombinedCriteria
boost::ptr_vector< Criteria >	mCriteriaList

Protected Attributes inherited from bayesopt::Criteria
NonParametricProcess *	mProc

randEngine *	mtRandom

Detailed Description

GP_Hedge model as describen in Hoffman et al.

[Hoffman2011]

The coefficients of the bandit algorithm has been carefully selected according to Shapire et al. Also, the implementation has been made to avoid over or underflow.

Definition at line 43 of file criteria_hedge.hpp.

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

Public Member Functions

Protected Member Functions

Protected Attributes

Detailed Description