m0_37600149的博客

  最近在看David Sliver强化学习的视频，只看视频不打代码会很无聊的：） 算法描述见上面的slides截图，下面直接贴代码   import numpy as np world_h = 5 world_w = 6 length = world_h * world_w gamma = 1 state = [i for i in range(length)] action...

AlphaGo论文： 　　AlphaGo:《Mastering the game of Go with deep neural networks and tree search》

AlphaGo Zero 论文:《Mastering the game of Go without human knowledge》 AlphaGo与AlphaGo Zero主要有以下几点不同： - AlphaGo中用了３个policy network,AlphaGo Zero只用了一个reinforcement learning network - AlphaGo Zero将po

Q-learning　　　DQN论文：Human-level control through deep reinforcement learning DQN其实就是将深度学习与Q-learning结合起来了，建立了卷积神经网络来估计Q值。 　　　　 　　建立了Q network,Q target network(Q̂ \hat Q)两个网络，对每一个episode,t时刻时，对于状态sts_t

https://www.jianshu.com/p/b9f14b8b1babhttps://github.com/openai/roboschool/issues/122