../
Online Opt HW3
1) A reduction from no-swap-regret to no-regret
Consider the Experts problem setting
There are some $N$ experts and the learner choose a mixed strategy over the experts at each turn.
| Learner | Adversary |
|---|---|
| 1. Pick $p_t\in\Delta_N$ | |
| Send $p_t$ | |
| 2. Choose $l_t\in[0,1]^N$ | |
| Send $l_t$ | |
| 3. Suffer loss $\langle p_t, l_t \rangle$ |