MATLAB Answers

Reinforcement Learning Simulink Block Inital Policy

1 view (last 30 days)
I want to give some initial policy to the reinforment learning block with DDPG agent. Not only the state and rewards but also the action needs to intialized how can it be done?

Answers (3)

Emmanouil Tzorakoleftherakis
If you already have a policy with trained weights, you could just use that directly when creating the agent, instead of creating a new network from scratch. Is this what you were looking for?

Syed Muhammad Yousaf Hashmy
I have a problem set which is hard one and for that I need to give the agent some initial policy as a hint i.e. a good initial starting point. So, how can I give the initial weights and then allow the agent to learn starting from good initial weights. How can I do that when I am using the rl block in simulink.

Emmanouil Tzorakoleftherakis
To use the rl agent block, you need to create an agent first, which also requires a policy architecture. When you set up your neural network, you can specify initial values for the weights using, e.g., the 'Weights' option of the fully connected layer (or any layer that has learnable parameters).

Community Treasure Hunt

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

Start Hunting!