# Reinforcement Learning
## DQN
離散Action 以DNN 模擬Q-table
why DQN cannot work on continues control?
reference: [DQN从入门到放弃7 连续控制DQN算法-NAF](https://zhuanlan.zhihu.com/p/21609472)

## Actor-Critic
Actor - Policy(π) Gradient
Actor_Loss
Critic - Q-Learning (Q^π(s,a))
Critic_Loss 
Critci網路算出Td-Error，利用Td-Error對Actor網路進行更新
目標—最小化Td-error

## TD(Temporal Difference)
why -Q(s,a)
reference: [時序差分學習](https://ithelp.ithome.com.tw/articles/10234455)

##