Skip to main navigation Skip to search Skip to main content

When Q-Learning fails: unstable behavior for infinite state spaces

Research output: Contribution to journalConference articlepeer-review

3 Downloads (Pure)

Abstract

The Q-learning algorithm is well known for its convergence guarantees to the optimal policy in finite-state environments. In this paper, we investigate its limitations in countable infinite state spaces – a setting common in real-world problems. To this end, we introduce a simple queueing model, based on a load balancing problem, with a countably infinite state space. In this model, a dispatcher assigns incoming jobs to one of two queues by choosing between two possible actions: ''red'' and ''green''. The ''red'' action leads to transient behavior, whereas the ''green'' action ensures stability. Our main result shows that, under certain parameter conditions, Q-learning exhibits instability and fails to converge to the optimal policy. Our findings reveal a critical gap in the theoretical understanding of model-free Reinforcement Learning methods in infinite domains. Numerical experiments illustrate that the transience also occurs with decreasing stepsizes that satisfy the usual Robbins-Monro conditions.
Original languageEnglish
Pages (from-to)79-83
Number of pages5
JournalPerformance Evaluation Review
Volume53
Issue number4
DOIs
Publication statusPublished - 31 Mar 2026
Event43rd International Symposium on Computer Performance, Modeling, Measurements and Evaluation 2025 - Amsterdam, Netherlands
Duration: 11 Nov 202513 Nov 2025

Fingerprint

Dive into the research topics of 'When Q-Learning fails: unstable behavior for infinite state spaces'. Together they form a unique fingerprint.

Cite this