This is the note for Berkeley CS285 Lecture 6. The lecture turns REINFORCE into actor-critic by learning a value function that gives each action a less noisy credit signal. The actor decides what to

This is the note for Berkeley CS285 Lecture 5. The lecture introduces the simplest model-free policy-gradient method: REINFORCE. The central idea is simple: make sampled actions more likely when thei

This is the note of Berkeley CS285, taught by Sergey Levine. Lecture 4 can be understood through one idea: A policy is good only if the future state distribution it creates leads to high reward. The

This is the note for Berkeley CS285 Lecture 3. The lecture asks how to make behavioral cloning work when a small prediction error can change the future data distribution. Strong imitation learning ne

This is the note of Berkeley CS285 taught by Sergey Levine. Lecture 1 introduces reinforcement learning (RL) as a framework for learning decisions from experience. The whole lecture can be understood

This is the note of Berkeley CS285 taught by Sergey Levine. Lecture 2 studies the simplest way to learn a policy from an expert: behavioral cloning. The central idea is easy to miss: Behavioral cloni

Welcome to Git! This is a quick, copy-paste friendly guide to get you productive fast. If you need deeper explanations, check the official Git docs or the excellent free book Pro Git. Quick StartInsta