DDPG Agent stopping training

2 次查看（过去 30 天）
Abd Al-Rahman Al-Remal
评论： Abd Al-Rahman Al-Remal 2021-7-1
Hi there,
When training my RL agent for some reason it stops trainign after ~100 episodes (usually less however) and the limit I have set is usually far less. I doubt it's the case that the training has stopped due to the 'isdone' signal in my simulink model (unless after only 100 episodes it managed to get exactly what I specified?).
Any help or advice would be much appreciated!
Edit: as I'm writing this I'm realising that part of my 'isdone' signal is for the output of my model to be excatly one value - could this be the cause as it obtains that value once and then stops?
  2 个评论
Abd Al-Rahman Al-Remal
Abd Al-Rahman Al-Remal 2021-6-17
Actually I think my value for the average reward to stop training was too low. I'll keep this updated if that works :)
Abd Al-Rahman Al-Remal
Abd Al-Rahman Al-Remal 2021-7-1
Yup this was the issue - all sorted now! :)

请先登录，再进行评论。

请先登录，再回答此问题。

回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Reinforcement Learning 的更多信息

标签

产品


版本

R2021a

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by