State of Algorithm 1
DefinitionFirstOrderFG_Supplement_Alg1_Statebandit-algorithmsgraph-theorymachine-learningp2o-gran-per-chapterp2o-plan-paperp2o-v1
The state of one Algorithm 1 instance consists of its fixed learning rate and its current EXP3 distribution on the ordinary probability simplex .
It is the state carried between played rounds; the sampling distribution is recomputed from each round.
Definition code
import Mathlib set_option autoImplicit false namespace FirstOrderFG.Supplement /-- **State of Algorithm 1 between rounds.** Zhang–Zhang–Luo–Mineiro, *Near-Optimal First-Order Regret for Bandits with Feedback Graphs*, arXiv:2610.40861, Algorithm 1, p.7: the fixed learning rate `η` and the current ordinary-simplex EXP3 distribution `p_t`. -/ structure Alg1.State (K : ℕ) where η : ℝ p : Fin K → ℝ end FirstOrderFG.Supplement
Source
Zhang–Zhang–Luo–Mineiro, Near-Optimal First-Order Regret for Bandits with Feedback Graphs, arXiv:2610.40861, Algorithm 1, p.7