25 #ifndef _CRITERIA_HEDGE_HPP_ 26 #define _CRITERIA_HEDGE_HPP_ 50 double operator() (
const vectord &x) {
return (*mCurrentCriterium)(x); };
52 bool requireComparison(){
return true; };
53 void initialCriteria();
54 bool rotateCriteria();
55 void pushResult(
const vectord& prevResult);
56 std::string getBestCriteria(vectord& best);
58 std::string name() {
return "cHedge";};
62 vectord loss_, gain_, prob_, cumprob_;
64 std::vector<vectord> mBestLists;
67 virtual double computeLoss(
const vectord& query)
68 {
return mProc->
prediction(query)->getMean(); }
79 std::string name() {
return "cHedgeRandom";};
82 virtual double computeLoss(
const vectord& query)
83 {
return mProc->prediction(query)->sample_query(); }
Namespace of the library interface.
Abstract class to implement Bayesian regressors.
GP_Hedge model as describen in Hoffman et al.
Abstract class for combined criteria functions.
Modification of the GP_Hedge algorithm where the bandit gains are random outcomes (like Thompson samp...
virtual ProbabilityDistribution * prediction(const vectord &query)=0
Function that returns the prediction of the GP for a query point in the hypercube [0...
Abstract interface for criteria functors.
Abstract module for combined criteria.