1
$\begingroup$

I'm working on a paper that uses ε-greedy algorithms for choosing episodes of a sarsa q-learning algorithms. I searched for algorithm but couldn't get so much. Can you please give me the algorithms and a little explanations about it?

$\endgroup$
  • 4
    $\begingroup$ This question asks to list a potentially large body of work. Community votes, please: is this too broad? $\endgroup$ – Raphael Feb 25 '16 at 17:24
  • $\begingroup$ @Raphael what you mean? what is the problem? $\endgroup$ – virtouso Feb 25 '16 at 18:19
  • $\begingroup$ @mtk99 "Working on", not "working with". It would be unusual to share a link to a paper that's not complete yet. $\endgroup$ – David Richerby Feb 25 '16 at 20:16
  • 3
    $\begingroup$ @DavidRicherby On the other hand, it is also unusual to ask strangers on the internet to explain the topic of the paper you are writing to you. $\endgroup$ – Raphael Feb 25 '16 at 21:00
  • $\begingroup$ Can you share a link to the paper? $\endgroup$ – Juan Leni Feb 25 '16 at 21:52
4
$\begingroup$

ε-greedy is just a way to promote exploration in Reinforcement Learning. I would not classify SARSA or Q-Learning as ε-greedy algorithms.

The latter are very common reinforcement learning algorithms and you can definitely can find lots of material about them.

I would suggest to have a look at this classic book (the draft for the second edition is available for free)

https://webdocs.cs.ualberta.ca/~sutton/book/the-book.html

Another shorter and excellent resource is (have a look at chapter 21):

Russell SJ, Norvig P, Davis E. Artificial intelligence: a modern approach. Upper Saddle River, NJ: Prentice Hall; 2010.

EDIT:

The paper you posted is using SARSA(λ), it is not the same as SARSA. You will need to go through a few Sutton chapters for that to make sense. You won't find it in Russell.

The link to SARSA(λ) in Sutton (1998) HTML version is here: webdocs.cs.ualberta.ca/~sutton/book/ebook/node77.html

However, I think you will need to read at least chapter 6 before jumping into eligibility traces

$\endgroup$
  • $\begingroup$ i think the paper meant sarsa from ε-greedy algorithm but im looking for a trace algorithm that how it really works $\endgroup$ – virtouso Feb 25 '16 at 16:10
  • $\begingroup$ This is the link to SARSA($\lambda$) in Sutton (1998) HTML version: webdocs.cs.ualberta.ca/~sutton/book/ebook/node77.html , however I think you will need to read at least chapter 6 before jumping into eligibility traces $\endgroup$ – Juan Leni Feb 25 '16 at 16:16
  • $\begingroup$ can you give just a simple trace example or explanation? the book just shows the algorithm. i cant understand it well. $\endgroup$ – virtouso Feb 25 '16 at 20:04

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.