datawhalechina / easy-rl Goto Github PK
View Code? Open in Web Editor NEW强化学习中文教程(蘑菇书🍄),在线阅读地址:https://datawhalechina.github.io/easy-rl/
License: Other
强化学习中文教程(蘑菇书🍄),在线阅读地址:https://datawhalechina.github.io/easy-rl/
License: Other
https://datawhalechina.github.io/easy-rl/#/
Description
第四章 策略梯度 有如下表述:
假设你在 st 执行 at,最后发现 τ 的奖励是正的, 那你就要增加这一项的概率。
反之,在 st 执行 at 会导致 τ 的奖励变成负的, 你就要减少这一项的概率。
我能明白这两句话是在表述我们的目的,但是公式中哪一部分在实现这个目的呢?
Thanks for this project.
An error is reported when running main.py in the A2C folder:
Start to train !
tensor([ 0.0307, 0.0015, -0.0309, -0.0313], device='cuda:0')
Traceback (most recent call last):
File "G:/My First Paper/Code/easy-rl/codes/A2C/main.py", line 105, in <module>
train(cfg,env,agent)
File "G:/My First Paper/Code/easy-rl/codes/A2C/main.py", line 64, in train
dist, value = agent.model(state)
File "D:\2019Download\Anaconda\Anaconda\envs\IL\lib\site-packages\torch\nn\modules\module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "G:/My First Paper/Code/easy-rl/codes\A2C\model.py", line 34, in forward
probs = self.actor(x)
File "D:\2019Download\Anaconda\Anaconda\envs\IL\lib\site-packages\torch\nn\modules\module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "D:\2019Download\Anaconda\Anaconda\envs\IL\lib\site-packages\torch\nn\modules\container.py", line 119, in forward
input = module(input)
File "D:\2019Download\Anaconda\Anaconda\envs\IL\lib\site-packages\torch\nn\modules\module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "D:\2019Download\Anaconda\Anaconda\envs\IL\lib\site-packages\torch\nn\modules\activation.py", line 1200, in forward
return F.softmax(input, self.dim, _stacklevel=5)
File "D:\2019Download\Anaconda\Anaconda\envs\IL\lib\site-packages\torch\nn\functional.py", line 1583, in softmax
ret = input.softmax(dim)
IndexError: Dimension out of range (expected to be in range of [-1, 0], but got 1)
Process finished with exit code 1
Looking forward to your guidance, thanks in advance.
在线笔记图太大字太小
https://datawhalechina.github.io/easy-rl/#/
Description
这里应该是叫自助采样,而非拔靴自助docs/chapter2/chapter2.md
https://github.com/datawhalechina/leedeeprl-notes/blob/0a9c1bd78d3d5d30f6c11c9d43e0e73863cadbf4/codes/double_dqn/dqn.py#L98-L108
这里使用的策略是Nature dqn而不是Double dqn吧?
第五章对应的习题中,“on-policy”和“off-policy”的区别。作者您只阐述了“on-policy”的描述和优劣势,并没有说到区分呀
https://datawhalechina.github.io/easy-rl/#/
Description
在 第四章 策略梯度 的 Tips 小节上面一点有如下表述:
对于 “TensorFlow 上调用现成函数”,我看到 SO 上有个 How to choose cross-entropy loss in TensorFlow? 这个问题,你建议 PG 这里怎么选呢?
https://datawhalechina.github.io/easy-rl/#/
Description
Lines 77 to 81 in e4690ac
On Line 78 and L79
example4.1 下面第一句话中:它的终止状态就是左上角跟『右上角 』 -->> 它的终止状态就是左上角跟『右下角 』
https://github.com/datawhalechina/leedeeprl-notes/blob/master/docs/chapter2/chapter2.md
公式(14)上面一段话中存在错字
使得每一次『得带』不需要更新所有的状态 ---->> 使得每一次『迭代』不需要更新所有的状态
在线笔记图太大字太小
你好,谢谢你们的项目,我学到了很多。
或许第二章节的Iterative Algorithm for Computing Value of a MRP这里的bootstrapping(拔靴自助)应该翻译为自举吗?
我参考的是Reinforcement Learning an introduction的翻译版强化学习(第二版)第87页第三段第三行的翻译,感觉这样顺口点,bootstrapping(拔靴自助)有点奇怪。
位置:https://github.com/datawhalechina/leedeeprl-notes/blob/master/docs/chapter2/chapter2_questions%26keywords.md
两者之间是由转换关系的
->
两者之间是有转换关系的
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.