Implementation of Markov Decision Processes into quantum algorithms for reinforcement learning

Abstract

In this work, we propose a methodology to implement classic Markov Decision Processes in a Quantum Computing paradigm, as a first step to achieve systems running Quantum Reinforcement Learning where both agent and environment are expressed as quantum programs. To do so, we analyze the interaction cycle between the agent and the environment in classic reinforcement learning and create a method to map a Markov Decision Process with discrete state space, action set, and rewards, into a quantum program.