You May Not Need Ratio Clipping in PPO
This addresses a key problem in reinforcement learning for practitioners by improving policy optimization methods, though it is incremental as it builds on PPO.
The paper tackles the issue of ratio clipping in Proximal Policy Optimization (PPO) by showing it can fail to bound ratios effectively, and proposes Early Stopping Policy Optimization (ESPO) that directly optimizes the original surrogate objective with early stopping, significantly outperforming PPO across many continuous control tasks.
Proximal Policy Optimization (PPO) methods learn a policy by iteratively performing multiple mini-batch optimization epochs of a surrogate objective with one set of sampled data. Ratio clipping PPO is a popular variant that clips the probability ratios between the target policy and the policy used to collect samples. Ratio clipping yields a pessimistic estimate of the original surrogate objective, and has been shown to be crucial for strong performance. We show in this paper that such ratio clipping may not be a good option as it can fail to effectively bound the ratios. Instead, one can directly optimize the original surrogate objective for multiple epochs; the key is to find a proper condition to early stop the optimization epoch in each iteration. Our theoretical analysis sheds light on how to determine when to stop the optimization epoch, and call the resulting algorithm Early Stopping Policy Optimization (ESPO). We compare ESPO with PPO across many continuous control tasks and show that ESPO significantly outperforms PPO. Furthermore, we show that ESPO can be easily scaled up to distributed training with many workers, delivering strong performance as well.