Per-arm regret of Algorithm 3
DefinitionFirstOrderFG_Supplement_Alg3_RegKbandit-algorithmsgraph-theorymachine-learningp2o-gran-per-chapterp2o-plan-paperp2o-v1
Against arm , Algorithm 3's expected regret is its expected incurred loss minus .
Theorem A.2 gives a first-order guarantee for this quantity simultaneously for every arm.
Definition code
import Mathlib
import Definitions.Def_FirstOrderFG_Supplement_Alg3_expLoss
import Definitions.Def_FirstOrderFG_Lk
set_option autoImplicit false
namespace FirstOrderFG.Supplement
/-- **Algorithm 3's expected regret against arm `k`.** Zhang–Zhang–Luo–Mineiro,
arXiv:2610.40861, Theorem A.2, p.15. -/
noncomputable def Alg3.RegK {K : ℕ} (E : Fin K → Fin K → Prop) [DecidableRel E]
(T : ℕ) (ℓ : ℕ → Fin K → ℝ) (k : Fin K) : ℝ :=
Alg3.expLoss E T ℓ - FirstOrderFG.Lk T ℓ k
end FirstOrderFG.Supplement
Source
Zhang–Zhang–Luo–Mineiro, Near-Optimal First-Order Regret for Bandits with Feedback Graphs, arXiv:2610.40861, Theorem A.2, p.15