Proximal Policy Optimization (PPO)

KL Divergence Penalty

Penalty added to PPO's objective function to control divergence between successive policies, adaptively adjusted to maintain updates within an acceptable region.

← Wstecz