PEA:Formulization The online learner(player)aims to make the prediction based by combining N experts'advice. At each round t=1,2,... (1)the player first picks a weight p from a simplex AN; (2)and simultaneously environments pick a loss vector eERN; (3)the player suffers loss fi(p:)(p,e),observes e and updates the model. The feasible domain is the(W-l)-dim simplex△v={p∈Rv|p,≥0，∑1pi=l} We typically assume that0≤lt,i≤1holds for all t∈[T]andi∈[W]. Advanced Optimization(Fall 2023) Lecture 6.Prediction with Expert Advice 7 Advanced Optimization (Fall 2023) Lecture 6. Prediction with Expert Advice 7 PEA: Formulization • The online learner (player) aims to make the prediction based by combining N experts’ advice