Per-arm regret of Algorithm 1
DefinitionFirstOrderFG_Supplement_Alg1_RegKbandit-algorithmsgraph-theorymachine-learningp2o-gran-per-chapterp2o-plan-paperp2o-v1
Against a fixed arm , Algorithm 1's expected regret is
This is the quantity bounded exactly in Theorem 3.2.
Definition code
import Mathlib
import Definitions.Def_FirstOrderFG_Supplement_Alg1_expLoss
import Definitions.Def_FirstOrderFG_Lk
set_option autoImplicit false
namespace FirstOrderFG.Supplement
/-- **Algorithm 1's expected regret against arm `k`.** Zhang–Zhang–Luo–Mineiro,
arXiv:2610.40861, §2, p.3, and Theorem 3.2, p.7. -/
noncomputable def Alg1.RegK {K : ℕ} (E : Fin K → Fin K → Prop) [DecidableRel E]
(T : ℕ) (ℓ : ℕ → Fin K → ℝ) (η : ℝ) (k : Fin K) : ℝ :=
Alg1.expLoss E T ℓ η - 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, §2, p.3; Theorem 3.2, p.7