Reinforcement Learning - Access agent/training parameters while training

Question

Tech Logg Ding 2021-8-4

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/892507-reinforcement-learning-access-agent-training-parameters-while-training

回答： Yash 2024-2-20

Hi,

I am writing to ask if there's anyway to access the agent or the training statistics during the training process?

For example, how can I extract the current reward of an episode in the training episode initialisation function?

Thank you.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Yash 2024-2-20

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/892507-reinforcement-learning-access-agent-training-parameters-while-training#answer_1412578

Normally, when you use the train function, it should open up the Reinforcement Learning Episode Manager. This tool allows you to see how the training is going. On the Episode Manager's chart, you can see the reward received for each training episode (EpisodeReward) and also the average reward value that's been calculated over a series of episodes (AverageReward).

Read more about the Episode Manager here: https://www.mathworks.com/help/reinforcement-learning/ug/train-reinforcement-learning-agents.html#:~:text=and%20GPUs.-,Episode%20Manager,-By%20default%2C%20calling

Hope this helps.