State-action-reward-state-action, or SARSA, is an algorithm for learning a Markov decision process policy, used in the reinforcement learning area of machine learning. It was proposed by Rummery and Niranjan under the name Modified Connectionist Q-Learning, and the alternative name SARSA, suggested by Richard Sutton, reflects the fact that its update rule for the Q-value depends on the agent's current state, the action it chooses, the reward it receives for that action, the state it enters afterward, and the next action it chooses in that new state.
Connections
Associated With
Contrasted directly in the reinforcement-learning literature as the standard off-policy/on-policy temporal-difference control pair. Verified en.wikipedia.org/wiki/State-action-reward-state-action: "While SARSA learns the Q values associated with taking the policy it follows itself, Watkin's Q-learning learns the Q values associated with taking the optimal policy while following an exploration/exploitation policy."
Reader Challenges (0)
No disputes yet. Spotted an error or a better source? Open the first one.
Sign in to dispute this or suggest a correction.