简体繁体 English

Microsoft CNTK加强学习C ++示例

[英]Microsoft CNTK reinforced learning C++ examples

原文 2017-06-05 20:11:43 9 1 c++/ neural-network/ reinforcement-learning/ cntk

I'm currently trying to write an AI in C++ that plays bomberman and I want to implement it as a neural network (reinforcement learning) with CNTK. 我目前正在尝试用C ++编写一个能够扮演轰炸机的AI，我想将其实现为CNTK的神经网络（强化学习）。 The toolkit has a lot of tutorials and examples, even one for reinforcement learning, but it is written in python, which I don't know. 该工具包有很多教程和示例，甚至是用于强化学习的教程和示例，但是它是用python编写的，我不知道。

https://github.com/Microsoft/CNTK/blob/v2.0.rc3/Tutorials/CNTK_203_Reinforcement_Learning_Basics.ipynb https://github.com/Microsoft/CNTK/blob/v2.0.rc3/Tutorials/CNTK_203_Reinforcement_Learning_Basics.ipynb

Are there any practical C++ examples or tutorials teaching reinforcement learning with CNTK? 是否有实用的C ++示例或使用CNTK进行强化学习的教程？

My first most issue is how to create a model with the CNTK C++ api. 我的第一个主要问题是如何使用CNTK C ++ API创建模型。

1 个解决方案

To create a model with the C++ API, take a look at these C++ examples . 要使用C ++ API创建模型，请查看以下C ++示例。

For reinforcement learning, you need an environment in which you can act and get rewards. 对于强化学习，您需要一个可以行动并获得奖励的环境。 The only environment I know out there in C++ is the Arcade Learning Environment . 我在C ++中唯一了解的环境是Arcade学习环境。 There are many environments available in Python such as openai's gym . Python中有许多可用的环境，例如openai的Gym 。

Unfortunately, CNTK doesn't have any examples of RL in C++. 不幸的是，CNTK没有C ++中的RL的任何示例。 If you follow the Python examples however and translate things in C++, there's no reason why you cannot make it work. 但是，如果您遵循Python示例并使用C ++进行翻译，则没有理由不能使其工作。