Q-learning
Christopher J. C. H. Watkins and Peter Dayan published Q-learning in Machine Learning in May 1992. The paper describes Q-learning as an incremental dynamic-programming method for agents to improve action-value estimates and presents a detailed convergence theorem for the method under stated conditions.
Historical classification
This verified record belongs to LINEAiGE's artificial-intelligence history layer beginning in 1956.
Why it matters
A documented reinforcement-learning milestone providing a formal published treatment and convergence result for Q-learning while explicitly identifying the method as originating in Watkins's earlier 1989 work.
People, organizations & technologies
PEOPLE
Christopher J. C. H. Watkins, Peter Dayan
ORGANIZATIONS
Machine Learning
TECHNOLOGIES
Q-learning, Reinforcement learning, Dynamic programming
Evidence & verification
STATE
verified
CONFIDENCE
high
LAST VERIFIED
2026-09-11T02:05:00+08:00
The Springer journal record verifies the authors, journal, volume, pages, May 1992 issue date, DOI, and abstract. The abstract explicitly states that Q-learning is an incremental method for dynamic programming, that it improves action-value evaluations, that the paper proves a convergence theorem, and that the method is attributed to Watkins (1989). Month precision is retained because the reviewed publisher record establishes May 1992 without an exact publication day. This record does not infer a 1989 thesis date beyond the year, development geography, ancestry, influence, or causal relationships.
Sources
Trace the chronology
Record identity
LINEAiGE IDq-learning-1992