SARSA
Definition
SARSA updates action values based on the actually following action.
An on-policy temporal-difference method that updates action values from the actually chosen subsequent action according to the current policy.
Simply explained
An agent learns from the current decision, its reward, and the next actually chosen step.
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 updates Q(s,a) from the reward and Q(s',a') of its next chosen action.
Limits and pitfalls
- Confusing SARSA with off-policy Q-learning or poorly managing exploration and learning rate.
Origin & history
Rummery and Niranjan described the method in 1994; the name SARSA became widespread later.
Use in mid-sized companies
- Possible example: An agent updates Q(s,a) from the reward and Q(s',a') of its next chosen action.
Frequently asked questions
What does SARSA mean?
An on-policy temporal-difference method that updates action values from the actually chosen subsequent action according to the current policy. What mistakes should be avoided? Confusing SARSA with off-policy Q-learning or poorly managing exploration and learning rate.
Sources
Source checked on 10.10.2026. Editorial application examples are hypothetical; no evidence of real customer results.
Editorially reviewed on: