How to train a Agent using reinforcement learning for variable step size ?

6 次查看（过去 30 天）
Chetan Thakur
Chetan Thakur 2024-1-18
编辑： Emmanouil Tzorakoleftherakis 2024-1-24
I have trained a Agent to perform certain task. But it works only for the step size used during training.
My qestion is, how can we train the Agent for variable step size. ?
e.g.
1) train the agent with step size of 1
2) next train the agent with step size of 10
... so on as required.
This could also mean how can we update agent with new learnings ? Without retraining on past dataset
  1 个评论
Emmanouil Tzorakoleftherakis
Emmanouil Tzorakoleftherakis 2024-1-24
编辑：Emmanouil Tzorakoleftherakis 2024-1-24
1) Can you clarify what you mean by variable step size? Are you referring to the solver options of the Simulink model? Or are you referring to the sample time of the agent?
2) If it's the latter, can you share a scenario/more details on the application? Typically when you train an agent it's for a specific task. Changing the sample time of the agent changes a lot of things. Presumably, similar tasks would have similar bandwidth requirements..

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

R2022b

Community Treasure Hunt

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

Start Hunting!

Translated by