behavior cloning

inverse reinforcement learning

reward shaping

curriculum learning

hierarchical RL

AC A2C A3C

pathwise derivative policy gradient

introduction of Q-learning

Tips of Q-learning

Q-learning for Continuous Actions

policy gradient

on-policy and off-policy

add constraint

ML 23-1 deep reinforcement learning

scenario of deep reinforcement learning

• learning to play GO

• Supervised vs Reinforcement

• applications

  Gym: https://gym.openai.com/

  Universe: https://openai.com/blog/universe/

• difficulties of reinforcement learning

  reward delay 一些没有奖励的动作在当前看起来没有用，但对未来会产生影响，帮助在未来得到奖励。

  agent’s actions affect the subsequent data it recevives，agent 需要去探索，不管是好的行为还是坏的。

• outline

Policy-based Approach - Learning an Actor

• machine learning $\approx$ looking for a function

• 找function 的三大步骤

  

• DRL

  1. neural network as actor

     input: vector、matrix，eg: pixels

     output: action 采取行动的几率，stochastic

  

  2. goodness of function

     supervised learning vs DRL

  
  3. pick the best
  • gradient ascent

  
  

  • add a baseline

critics

评估observation

Actor-Critic

ML 23-2 policy gradient (Supplementary Explanation)

ML 23-3 RL

interact with environments

机器学到的行为会影响下一步的发展，所有的action 当成整体看待

components

env、reward function不能控制，只能调整actor的行为

critic

评估critic：

Monre-Carlo：

Temporal defference：

Q

actor 如果⽆法穷举则会爆炸，采用PDPG

pathwise derivative policy gradient

Asynchronous A3C

imitation learning

类似GAN: