BayesOpt
|
Posterior model of nonparametric processes/criteria based on MCMC samples. More...
#include <posterior_mcmc.hpp>
Public Types | |
typedef boost::ptr_vector< NonParametricProcess > | GPVect |
typedef boost::ptr_vector< Criteria > | CritVect |
Public Member Functions | |
MCMCModel (size_t dim, Parameters params, randEngine &eng) | |
Constructor (Note: default constructor is private) More... | |
void | updateHyperParameters () |
void | fitSurrogateModel () |
void | updateSurrogateModel () |
double | evaluateCriteria (const vectord &query) |
void | updateCriteria (const vectord &query) |
bool | criteriaRequiresComparison () |
void | setFirstCriterium () |
bool | setNextCriterium (const vectord &prevResult) |
std::string | getBestCriteria (vectord &best) |
ProbabilityDistribution * | getPrediction (const vectord &query) |
Public Member Functions inherited from bayesopt::PosteriorModel | |
PosteriorModel (size_t dim, Parameters params, randEngine &eng) | |
Constructor. More... | |
virtual | ~PosteriorModel () |
Default destructor. | |
void | setSamples (const matrixd &x, const vectord &y) |
void | setSamples (const matrixd &x) |
void | setSamples (const vectord &y) |
void | setSample (const vectord &x, double y) |
void | addSample (const vectord &x, double y) |
double | getValueAtMinimum () |
vectord | getPointAtMinimum () |
void | plotDataset (TLogLevel level) |
const Dataset * | getData () |
Private Member Functions | |
void | setSurrogateModel (randEngine &eng) |
void | setCriteria (randEngine &eng) |
MCMCModel (MCMCModel ©) | |
Private Attributes | |
size_t | nParticles |
GPVect | mGP |
Pointer to surrogate model. | |
CritVect | mCrit |
Metacriteria model. | |
boost::scoped_ptr< MCMCSampler > | kSampler |
Additional Inherited Members | |
Static Public Member Functions inherited from bayesopt::PosteriorModel | |
static PosteriorModel * | create (size_t dim, Parameters params, randEngine &eng) |
Protected Attributes inherited from bayesopt::PosteriorModel | |
Parameters | mParameters |
Configuration parameters. | |
size_t | mDims |
Number of dimensions. | |
Dataset | mData |
Dataset (x-> inputs, y-> labels/output) | |
MeanModel | mMean |
Posterior model of nonparametric processes/criteria based on MCMC samples.
For computational reasons we store a copy of each conditional models with the corresponding particle generated by MCMC. That is to avoid costly operations like matrix inversions for every kernel parameter in a GP prediction. Thus, we assume that the number of particles is not very large.
Definition at line 47 of file posterior_mcmc.hpp.
bayesopt::MCMCModel::MCMCModel | ( | size_t | dim, |
Parameters | params, | ||
randEngine & | eng | ||
) |
Constructor (Note: default constructor is private)
dim | number of input dimensions |
params | configuration parameters (see parameters.hpp) |
eng | random number generation engine (boost) |
Definition at line 27 of file posterior_mcmc.cpp.
References mGP.