Policy Advantage Function

In reinforcement learning, a function measuring how much better or worse taking a given action in a given state is relative to average. Positive advantage encourages the action, while negative advantage suppresses it. It is a core component of policy gradient methods.