Reinforcement Learning – A Reference

Problem: SARSA pushes q-values towards the current policy, but ideally we'd want optimal values. Solution: Use the best action in TD-target calculation -> Q-learning

Perhaps someone else will find it helpful!