Online Learning:Formulation At each round t=l,2,· (1)the player first picks a model wtW; (2)and simultaneously environments pick an online function fr:W->R; (3)the player suffers loss fi(wt),observes some information about fi and updates the model. Spam filtering (1)Player submits a spam classifier w ↓ (2)A mail is revealed whether it is a spam ☒ (3)Player suffers loss fi(w)and updates model Advanced Optimization(Fall 2023) Lecture 5.Online Convex Optimization 9 Advanced Optimization (Fall 2023) Lecture 5. Online Convex Optimization 9 Online Learning: Formulation Spam filtering