1
$\begingroup$

Take a n-armed bandit setup. When agent pulls an arm, the environment reveals rewards of all arms at that time step. My question is will existing algorithms work well in this setting ? Can we take advantage of all the information revealed to us ?

$\endgroup$

1 Answer 1

1
$\begingroup$

Yes, I think many algorithms take the advantage of all the information revealed by all the other arms at a time stamp.

Let's take RL (Reinforcement Learning) for e.g

There is a trade off of exploration and exploitation in RL. We apply e-greedy policy in which our agent chooses the action that corresponds to the largest expected value with e probability. In this we can easily see our agent is exploiting information given by different arms to continue to learn more about them.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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