Update Discrete Action Values of a Q-Agent in Reinforcement Learning each Episode

Hello,

I'm working on a RL project where I have two discrete actions for my Q-Learning QAgent, and I want to know if the set values can be change or not, in the project that I need to do I was planning to update the discrete action values each run episode and used them to feed a PI controller. So someone know if once the values are set can be update and changed for different ones or not?

Answer 1

[Saksham]

Hi Ander,

I understand that you wish to know if set values for Discrete Action Values can be changed per episode or not.

You may find ‘rlFiniteSetSpec’ helpful for the use case. Below is the link to its documentation:

https://in.mathworks.com/help/reinforcement-learning/ref/rl.util.rlfinitesetspec.html

Sincerely,

Saksham