jia
New Member
Posts: 5
|
Post by jia on Oct 15, 2020 15:13:58 GMT
So this exercise asks us to prove that swap regret of an action sequence of length T can exceed its external regret by at least T.
It's easy to show that the total swap regret can exceed the total external regret by at least T if the player also swaps to the winning side. However, since both regrets are defined as average regret, it seems impossible to get a total difference of T^2 and then get averaged difference of T. Is this a typo or am i missing something?
Thanks!
|
|