Gradient in RL DDPG Agent

2 visualizaciones (últimos 30 días)
Prashanth Chivkula
Prashanth Chivkula el 9 de Mayo de 2021
Where can I find the function description for the gradient while finding the critic gradient or the actor gradient used in RL DDPG agent? When I open gradient by right clicking on it , I get the Numerical gradient help file, Is this the correct one? What does the text 'output-input' mean when the dQdInput is taken in the next line of the code.
I want to use the gradient function for my custom agent.
criticGradient = gradient(this.Critic,'loss-parameters',...
[Observations, Actions], targetq);
dQdInput=gradient(obj.Critic,'output-input',[Observations,Actions]);

Respuesta aceptada

Emmanouil Tzorakoleftherakis
Emmanouil Tzorakoleftherakis el 13 de Mayo de 2021
If you put a break point right before 'gradient' is called in this example, you can step in and see the function implementation. This method is not 'public', that's why there is no formal doc page, but you can see the description below.
Hope that helps

Más respuestas (0)

Community Treasure Hunt

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

Start Hunting!

Translated by