QSoftmaxPolicy.hpp

Go to the documentation of this file.

#ifndef AI_TOOLBOX_MDP_Q_SOFTMAX_POLICY_HEADER_FILE
#define AI_TOOLBOX_MDP_Q_SOFTMAX_POLICY_HEADER_FILE
 
#include <AIToolbox/MDP/Policies/QPolicyInterface.hpp>
#include <AIToolbox/MDP/Policies/QGreedyPolicy.hpp>
 
namespace AIToolbox::MDP {
    class QSoftmaxPolicy : public QPolicyInterface {
        public:
            QSoftmaxPolicy(const QFunction & q, double temperature = 1.0);
 
            virtual size_t sampleAction(const size_t & s) const override;
 
            virtual double getActionProbability(const size_t & s, const size_t & a) const override;
 
            virtual Matrix2D getPolicy() const override;
 
            void setTemperature(double t);
 
            double getTemperature() const;
 
        private:
            double temperature_;
            // To avoid reallocating a vector every time for sampling.
            mutable std::vector<size_t> bestActions_;
            mutable Vector vbuffer_;
    };
}
 
#endif