1. Reinforcement Learning

  • Actor(Policy)

    Neural Network as Actor (Deep). vs lookup Table(Q Learning).

使用神经网络作为Actor比查表的优势?

查表无法穷举输入,e.g.图像画面或者语言输入。NN泛化性比较强,对于未看过的Observation,举一反三,合理的输出。
  • Environment

  • Reward

2. Deep Learning

  • 如何选取Actor? Neural Network as Actor (Deep)

  • 如何衡量Actor的好坏?

    Maxmize Reward的期望。Reward是一个回合episode,每轮Reward的总和。由于Actor是stochastic随机的,每个回合的Reward不同。所以maxmize sampling N回合Reward的期望。

    期望就衡量了Actor

actor_goodness

Share on: TwitterFacebookEmail

Comments


Related Posts


Published

Category

NLP

Tags

Contact