Unlock: Q-Learning

Model-free, off-policy value learning: the Q-learning update rule, convergence under Robbins-Monro conditions, and the deep Q-network revolution that introduced function approximation, experience replay, and the deadly triad.