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 condition

Stats

• citations0
• S2 citations—
• github stars0
• HF likes0
• heat score0.00
• arxiv keyhazan2018provably

Related papers

• The Importance Of Non-markovianity In Maximum State Entropy Exploration (2022)0.00
• Fast Rates For Maximum Entropy Exploration (2023)0.00
• Maximum Entropy Exploration Without The Rollouts (2026)0.00
• Task-agnostic Exploration Via Policy Gradient Of A Non-parametric State Entropy Estimate (2020)0.00
• Maximum-entropy Exploration With Future State-action Visitation Measures (2026)0.00
• Parameterized Mdps And Reinforcement Learning Problems -- A Maximum Entropy Principle Based Framework (2020)8.60
• Off-policy Maximum Entropy RL With Future State And Action Visitation Measures (2024)0.00
• Improved Bounds For Reward-agnostic And Reward-free Exploration (2026)0.00