Keavnn

icon indicating a website link https://stepneverstop.github.io icon indicating an email [email protected]

icon location Nan Jing icon location Ph.D. candidate at Lamda 5 group in Nanjing University. My research interest is deep reinforcement learning.

Results 7 comments of Keavnn

实现新的强化学习算法

- [ ] 优化MARL中的训练部分，避免繁多的键值索引

Broken pipe

@GongXiaomeng 你的启动命令是什么

Broken pipe

@GongXiaomeng 我无法复现该问题，你可以详细说说，或者日后我添加对`robosuite `的支持，并做测试。

通用的经验表示格式，结合经验池机制优化数据流

- 适配gym - 测试LSTM是否正确开启训练 - 修复On-policy算法

masac

This repo is under reconstruction right now, so the code implementation in the repository is not rigorously tested.

Check that the code implementation is accurate and reasonable

- [x] 检查将代码中关于运算维度的选择(dim/axis)把能设置为-1的都设置为-1。

Check that the code implementation is accurate and reasonable

- [x] 校正RNN隐状态在使用探索策略时的迭代更新 [abf6b0a](https://github.com/StepNeverStop/RLs/commit/abf6b0aee4ccbe67c2da1c981117c396bec25843) - [x] 实现按策略与环境交互的间隔更新策略 [abf6b0a](https://github.com/StepNeverStop/RLs/commit/abf6b0aee4ccbe67c2da1c981117c396bec25843)