Best-arm regret of Algorithm 2
DefinitionFirstOrderFG_Supplement_Alg2_Regbandit-algorithmsgraph-theorymachine-learningp2o-gran-per-chapterp2o-plan-paperp2o-v1
The published Algorithm 2 instance's best-arm expected regret is its expected incurred loss minus .
This is the regret appearing after the tuned choice of in Theorem 4.2.
Definition code
import Mathlib
import Definitions.Def_FirstOrderFG_Supplement_Alg2_expLoss
import Definitions.Def_FirstOrderFG_Lstar
set_option autoImplicit false
namespace FirstOrderFG.Supplement
/-- **Published Algorithm 2's expected best-arm regret.**
Zhang–Zhang–Luo–Mineiro, arXiv:2610.40861, §2, p.3, and Theorem 4.2, p.11. -/
noncomputable def Alg2.Reg {K : ℕ} (E : Fin K → Fin K → Prop) [DecidableRel E]
(T : ℕ) (ℓ : ℕ → Fin K → ℝ) (q₀ : (Fin K → ℝ) → ℝ → Fin K → ℝ)
(A θ : ℝ) : ℝ :=
Alg2.expLoss E T ℓ q₀ A θ - FirstOrderFG.Lstar T ℓ
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 4.2, p.11