Model-Free Reinforcement Learning

A reinforcement learning approach that does not explicitly learn an environment model, instead estimating value functions or policy gradients directly from samples gathered through environment interaction. It typically has lower sample efficiency but is simpler to implement.