Q-Learning
Definition
Q-learning learns action values from observed transitions and rewards.
A model-free off-policy temporal-difference method that estimates the optimal Q-function via a Bellman optimality update from observed transitions.
Simply explained
An agent tries out actions and improves its estimate of which decisions pay off in the long run.
Why it matters in practice
The term helps to select, develop, evaluate AI systems in a well-founded manner and realistically assess their technical risks.
Example from business
An agent learns a decision policy from states, actions, rewards, and subsequent states.
Limits and pitfalls
- Assuming convergence without the theoretically necessary conditions such as sufficient exploration, appropriate learning rates, and a suitable stationary problem, or using function approximation without verification.
Origin & history
Watkins developed Q-learning in 1989; Watkins and Dayan proved convergence under conditions in 1992.
Use in mid-sized companies
- Possible application: An agent learns a decision policy from states, actions, rewards, and subsequent states.
Frequently asked questions
What does Q-learning mean?
A model-free off-policy temporal-difference method that estimates the optimal Q-function via a Bellman optimality update from observed transitions. What mistakes should be avoided? Assuming convergence without the theoretically necessary conditions such as sufficient exploration, appropriate learning rates, and a suitable stationary problem, or using function approximation without verification.
Sources
Source checked on 10.10.2026. Editorial application examples are hypothetical; no evidence of real customer results.
Editorially reviewed on: