backup operator

expected-max q-learning (1) q-learning operator (1) behavioral policy (1) d4rl benchmark (1) behavior policy (1) off-policy learning (1) value iteration (1) value function (1) offline reinforcement learning (1) sample efficiency (1)