Intro

这篇blog是我在看过Logan等人的“implementation matters in deep policy gradients: a case study on ppo and trpo“之后的总结。

value function clipping

reward scaling

orthogonal initialization and layer scaling

adam learning rate annealing

reward clipping

observation clipping

hyperbolic tan activations

global gradient clipping

针对PPO的一些Code-level性能优化技巧

原文：https://www.cnblogs.com/dynmi/p/14031724.html