Nature

Relative importance sampling for off-policy actor-critic in deep reinforcement learning

Figure 1a illustrates that off-policy learning primarily involves two policies: the behavioral policy (b), also known as the sampling distribution, and the target policy (\(\pi\)), also known as the ...
Nature

Risk sensitive twin distributional critics with a lambda lower confidence bound for continuous control reinforcement learning

Off-policy actor–critic methods such as Twin Delayed Deep Deterministic Policy Gradient (TD3) are the workhorse of continuous-control reinforcement learning. However, they rely on scalar value ...