Post by jia on Oct 13, 2020 16:09:43 GMT
This problem asks us to consider a thought experiment in which we take into account the current cost vector.
I followed the hint but not sure if I understood it correctly. From my understanding, the hint suggests we first consider the best action in hindsight. Then we work backward from time T, and iteratively transform the best action in hindsight to the action proposed by the algorithm. Is this correct?
If yes, we are supposed to compare the cost difference between two actions at each transform and find that the cost only decreases with each step of transformation. I guess the cost refers to the sum of the untransformed best actions and the transformed actions. However, how the cost decreases with each step of transformation is not clear to me. Can you provide more hints on this?
Thanks a lot!
I followed the hint but not sure if I understood it correctly. From my understanding, the hint suggests we first consider the best action in hindsight. Then we work backward from time T, and iteratively transform the best action in hindsight to the action proposed by the algorithm. Is this correct?
If yes, we are supposed to compare the cost difference between two actions at each transform and find that the cost only decreases with each step of transformation. I guess the cost refers to the sum of the untransformed best actions and the transformed actions. However, how the cost decreases with each step of transformation is not clear to me. Can you provide more hints on this?
Thanks a lot!
