一、模拟环境:
1)OpenAI提供的Gym和Universe(可以看成是Gym(https://www.gymlibrary.ml/)的升级版本,包含更多更复杂的环境,比如PC游戏);
2)DeepMind Lab提供的3D游戏环境;
3)Mata(原Facebook)公司提供实时策略游戏的FAIR TorchCraft;
4)ViZDoom是3D射击游戏Doom的测试平台;
5)TORCS是一个3D赛车模拟环境;
6)MuJoCo是一个物理模拟环境,支持连续性动作空间类的任务,是收费的,不过Gym集成了。
二、组合优化
《Reinforcement Learning for Combinatorial Optimization: A Survey》
三、多智能体
多智能体之间的关系类型,包括完全合作式、完全竞争式和混合关系式。
相关论文:
《Multi-Agent Reinforcement Learning: A Selective Overview of Theories and Algorithms》
《A comprehensive survey of multi-agent reinforcement learning》
《Littman M L. Markov games as a framework for multi-agent reinforcement learning[C]. international conference on machine learning》
《The dynamics of reinforcement learning in cooperative multiagent systems》
《Mean Field Multi-Agent Reinforcement Learning[C]. international conference on machine learning》
四、强化学习综述
《Deep Reinforcement Learning: An Overview》
https://arxiv.org/abs/1701.07274
五、强化学习的相关课程
https://www.udacity.com/course/reinforcement-learning—ud600(Georgia Tech, CS 8803)
http://web.stanford.edu/class/cs234/index.html(Stanford, CS234)
http://rll.berkeley.edu/deeprlcourse/(Berkeley, CS 294, Fall 2017)
https://www.udemy.com/deep-reinforcement-learning-in-python/(Udemy高级教程)
https://blog.csdn.net/qq_36829091/article/details/83213707(李宏毅老师视频讲解)
六、强化学习的框架与代码实现
6.1)tensorflow在强化学习算法的实现:https://github.com/reinforceio/tensorforce
6.2)深度强化学习实验室github地址:https://github.com/NeuronDance/DeepRL
6.3)强化学习框架Deep Learning and Reinforcement Learning Library for Scientists:https://github.com/tensorlayer/tensorlayer
6.4)强化学习实现代码:https://github.com/Teacher-Guo/RL_code
6.5)清华大学强化学习开源框架-天授:https://github.com/thu-ml/tianshou
6.6)Rainbow:整合DQN六种改进的深度强化学习方法!:https://github.com/princewen/tensorflow_practice/tree/master/RL/Basic-Rainbow-Net
七、强化学习在NLP的应用
https://www.microsoft.com/en-us/research/wp-content/uploads/2017/11/zhang.pdf
https://www.zhihu.com/question/47548097