P
Initializing...
Eq. (7) — the dual iterates of Algorithm 1 have average regret at most $GD/\sqrt T\le\epsilon$ · Prove2Me