Retrace(λ)

Safe and Efficient Off-Policy Reinforcement Learning