Entropy Regularized Reinforcement Learning Using Large Deviation Theory

Abstract

Reinforcement learning (RL) is an important field of research in machine learning that is increasingly being applied to complex optimization problems in physics. In parallel, concepts from physics have contributed to important advances in RL with developments such as entropy-regularized RL. While these developments have led to advances in both fields, obtaining analytical solutions for optimization in entropy-regularized RL is currently an open problem. In this paper, we establish a mapping between entropy-regularized RL and research in non-equilibrium statistical mechanics focusing on Markovian processes conditioned on rare events. In the long-time limit, we apply approaches from large deviation theory to derive exact analytical results for the optimal policy and optimal dynamics in Markov Decision Process (MDP) models of reinforcement learning. The results obtained lead to a novel analytical and computational framework for entropy-regularized RL which is validated by simulations. The mapping established in this work connects current research in reinforcement learning and non-equilibrium statistical mechanics, thereby opening new avenues for the application of analytical and computational approaches from one field to cutting-edge problems in the other.

References in corpus (8)

• First-order dynamical phase transition in models of glasses: an approach based on ensembles of histories
• Machine learning meets quantum physics
• Construction of a Coordinate Bethe Ansatz for the asymmetric simple exclusion process with open boundaries
• Measurement Based Feedback Quantum Control With Deep Reinforcement Learning for Double-well Non-linear Potential
• Adaptive Optics control using Model-Based Reinforcement Learning
• Reinforcement learning of rare diffusive dynamics
• Properties of a non-equilibrium heat bath
• Physics-informed graph neural networks enhance scalability of variational nonequilibrium optimal control