I’ve looked at various articles and I’m still very confused, I understand the normal double Q learning about having two Action value estimates that use two different set of samples

But coming to neural networks I’m confused The normal DQN algorithm uses our target network for both action selection and evaluation when performing updates

I was told that they initialize them to small weights to prevent overestimation, if that’s the case why exactly are we now using the online network to select actions

I mean our target network can handle that since it’ll never overestimate values

Can someone shed more light on this?

Thank you in advance!

  • 2
    $\begingroup$ I have already answered your question here: ai.stackexchange.com/questions/22787/… and tried to correct your misunderstanding about the target network . . . could you please comment on the answer about which part of my answer you did not understand. Here you are asking almost the same question again, and have not given any feedback on existing Q&A $\endgroup$ – Neil Slater Aug 1 '20 at 11:24
  • $\begingroup$ @Neil slater I’ll check out the answer right away $\endgroup$ – Chukwudi Aug 1 '20 at 18:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.