Provably Efficient Maximum Entropy Exploration

Abstract

Suppose an agent is in a (possibly unknown) Markov Decision Process in the absence of a reward signal, what might we hope that an agent can efficiently learn to do? This work studies a broad class of objectives that are defined solely as functions of the state-visitation frequencies that are induced by how the agent behaves. For example, one natural, intrinsically defined, objective problem is for the agent to learn a policy which induces a distribution over state space that is as uniform as possible, which can be measured in an entropic sense. We provide an efficient algorithm to optimize such such intrinsically defined objectives, when given access to a black box planning oracle (which is robust to function approximation). Furthermore, when restricted to the tabular setting where we have sample based access to the MDP, our proposed algorithm is provably efficient, both in terms of its sample and computational complexities. Key to our algorithmic methodology is utilizing the conditional gradient method (a.k.a. the Frank-Wolfe algorithm) which utilizes an approximate MDP solver.

Updated experiment results; minor revisions in writing

Cited by in corpus (10)

• FLAMBE: Structural Complexity and Representation Learning of Low Rank MDPs
• On Reward-Free Reinforcement Learning with Linear Function Approximation
• Learning latent state representation for speeding up exploration
• Exploration-Exploitation Trade-off in Reinforcement Learning on Online Markov Decision Processes with Global Concave Rewards
• Entropy Regularization with Discounted Future State Distribution in Policy Gradient Methods
• Marginalized State Distribution Entropy Regularization in Policy Optimization
• Autonomous exploration for navigating in non-stationary CMPs
• Hamilton-Jacobi-Bellman Equations for Maximum Entropy Optimal Control
• A maximum-entropy approach to off-policy evaluation in average-reward MDPs
• Adaptive Sampling for Estimating Multiple Probability Distributions