How can I create a Reinforcement Learning Agent representation based on Recurrent neural network (RNN, LSTM, among others)

Dear all,

How can I create a Reinforcement Learning Agent representation based on Recurrent neural network (RNN, LSTM, among others).

Is it possible??

Thanks

Answer 1

[Emmanouil Tzorakoleftherakis]

Hello,

Which release are you using? R2020a and R2020b support LSTM policies for PPO and DQN agents. Starting in R2021a you can additionally use PG, DDPG, AC, SAC, and TD3 agents with LSTMs.

Some resources and examples that may be helpful:

https://www.mathworks.com/help/reinforcement-learning/ug/create-policy-and-value-function-representations.html#mw_7822839a-3568-43af-a6f1-1920d89ccbdb

https://www.mathworks.com/help/releases/R2020a/reinforcement-learning/ref/rldqnagent.html#mw_c74efdf2-bac8-42c1-941b-6ec34b04f3a6

https://www.mathworks.com/help/releases/R2020a/reinforcement-learning/ref/rlppoagent.html#mw_d79da4e4-a3f2-4baf-8d2d-b2ded25baa6b

https://www.mathworks.com/help/reinforcement-learning/ug/water-distribution-scheduling-system.html

Hope that helps

[Juan Vasconez]

Thanks a lot.! That solved my issue.!

[dong]

Hello, how to find recursive network module in simulink, I used version 2019a.