Geometry-Inspired Unified Framework for Discounted and Average Reward MDPs

2025-10-27

AI Summary

This paper unifies the theoretical analysis of discounted and average-reward Markov decision processes (MDPs). Addressing the long-standing theoretical separation between these two MDP classes, it extends the geometric perspective of discounted MDPs to the average-reward setting, establishing a unified modeling framework grounded in the geometric structure of the state space. By integrating dynamic programming, ergodicity theory, and value iteration, the authors prove that, under the assumption of a unique ergodic optimal policy, value iteration exhibits geometric convergence in both discounted and average-reward MDPs. This work achieves a deep theoretical unification of the two paradigms and, for the first time, establishes geometric convergence of value iteration in average-reward MDPs—providing a novel analytical paradigm for reinforcement learning and optimal control.

Technology Category

• Reasoning under Uncertainty: Sequential Decision Making
• Game Theory and Economic Paradigms: Adversarial Learning
• Planning, Routing, and Scheduling: Planning with Markov Models (MDPs, POMDPs)

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems

Abstract

The theoretical analysis of Markov Decision Processes (MDPs) is commonly split into two cases - the average-reward case and the discounted-reward case - which, while sharing similarities, are typically analyzed separately. In this work, we extend a recently introduced geometric interpretation of MDPs for the discounted-reward case to the average-reward case, thereby unifying both. This allows us to extend a major result known for the discounted-reward case to the average-reward case: under a unique and ergodic optimal policy, the Value Iteration algorithm achieves a geometric convergence rate.

Problem

• Unifying discounted and average reward MDP analysis
• Extending geometric interpretation to average-reward MDPs
• Establishing geometric convergence for Value Iteration

Innovation

• Unified geometric framework for MDPs
• Extends discounted results to average rewards
• Proves geometric convergence for Value Iteration