QSoftmaxPolicy.hpp

Go to the documentation of this file.

#ifndef AI_TOOLBOX_BANDIT_Q_SOFTMAX_POLICY_HEADER_FILE
#define AI_TOOLBOX_BANDIT_Q_SOFTMAX_POLICY_HEADER_FILE
 
#include <AIToolbox/Bandit/Types.hpp>
#include <AIToolbox/Bandit/Policies/PolicyInterface.hpp>
 
namespace AIToolbox::Bandit {
    class QSoftmaxPolicy : public PolicyInterface {
        public:
            QSoftmaxPolicy(const QFunction & q, double temperature = 1.0);
 
            virtual size_t sampleAction() const override;
 
            virtual double getActionProbability(const size_t & a) const override;
 
            void setTemperature(double t);
 
            double getTemperature() const;
 
            virtual Vector getPolicy() const override;
 
        private:
            double temperature_;
            const QFunction & q_;
            // To avoid reallocating a vector every time for sampling.
            mutable std::vector<size_t> bestActions_;
            mutable Vector vbuffer_;
    };
}
 
#endif