Optimal Non-Asymptotic Rates of Value Iteration for Average-Reward Markov Decision Processes

Authors: Jongmin Lee, Ernest Ryu

ICLR 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical In this work, we conduct refined non-asymptotic analyses of average-reward MDPs, obtaining a collection of convergence results that advance our understanding of the setup. Among our new results, most notable are the O(1/k)-rates of Anchored Value Iteration on the Bellman error under the multichain setup and the span-based complexity lower bound that matches the O(1/k) upper bound up to a constant factor of 8 in the weakly communicating and unichain setups.
Researcher Affiliation Academia Jongmin Lee Seoul National University Department of Mathematical Sciences EMAIL Ernest K. Ryu UCLA Department of Mathematics EMAIL
Pseudocode No The Relaxed Value Iteration (Rx-VI) is V k = λk V k 1 + (1 λk)TV k 1 (Rx-VI) for k = 1, 2, . . . , where T is the Bellman optimality operator, V 0 Rn is a starting point, and 0 λk < 1 for k = 0, 1, . . . . πk is a greedy policy satisfying T πk V k = TV k for k = 0, 1, . . . . The Anchored Value Iteration is V k = λk V 0 + (1 λk)TV k 1 (Anc-VI) for k = 1, 2, . . . , where T is the Bellman optimality operator, V 0 Rn is a starting point, and 0 λk < 1 for k = 0, 1, . . . .
Open Source Code No No explicit statement about code release, repository link, or code in supplementary materials for the methodology described in this paper was found.
Open Datasets No The paper focuses on theoretical analysis of average-reward MDPs and does not report on experiments using specific datasets, therefore no dataset access information is provided.
Dataset Splits No The paper is theoretical and does not involve empirical experiments with datasets; therefore, no dataset splits are discussed.
Hardware Specification No The paper presents theoretical results and does not describe any experimental setup or specific hardware used for computations.
Software Dependencies No The paper is purely theoretical, focusing on mathematical analysis and proofs, and does not mention any software dependencies with version numbers for experimental replication.
Experiment Setup No The paper is theoretical and does not include an experimental section or describe any specific experimental setup, hyperparameters, or training configurations.