Scalar confidence on the complete terminal history
ProvedLG2021.Concentration.estimatorGood_full_scalarFix , , a compact convex legal set with pairwise diameter bound , and the published configuration. Assume a one-Lipschitz loss with unit range on the legal set and the normalized adaptive noise condition. Let and . For a complete -observation history, suppose all candidate norms obey the configured bound and its first observations lie in the published event . Then every scalar estimate at every healthy actual trace state and every candidate differs from its true surrogate gap by at most the published confidence radius at the actual sample count.
This is a deterministic implication about the complete terminal history. It connects the paired-prefix confidence event to the estimates used by the actual policy. The probability cost of that event is supplied separately by the simultaneous estimator-failure theorem.
This is a formalization auxiliary for Algorithms 1–2 and their probability analysis in Tor Lattimore and András György, Improved Regret for Zeroth-Order Stochastic Convex Bandits, COLT 2021, PMLR 134, Lemma 4 and Sections 5–7. The implementation and event predicates are published definitions; the guarantee is proved in this theorem. Adaptive observation kernels and exact finite-history semantics are explicit formal interfaces. Statistical concentration and regret claims are not assumed by the implementation definitions.
import Mathlib import Definitions.Def_LG2021_Policy open LG2021 LG2021.Algorithm LG2021.Concentration BilateralProfitMHR MeasureTheory ProbabilityTheory Set Metric
theorem LG2021.Concentration.estimatorGood_full_scalar {d n u N : ℕ} (hd : 0 < d) (hn : 0 < n)
{D : ℝ} (hD : 0 ≤ D) {K : Set (Space d)} (hK : IsCompact K) (hcv : Convex ℝ K)
(hdiam : ∀ x ∈ K, ∀ y ∈ K, dist x y ≤ D) (R : ℝ) (anchor : Space d) (ha : anchor ∈ K)
(ν : ObsKernel (Space d)) (hνm : ν.IsMeasurable) (F : Space d → ℝ)
(hLip : LipschitzWith 1 F) (hF : ∀ x ∈ K, F x ∈ Set.Icc (0 : ℝ) 1)
(hν : CorollaryTransport.UnitNoiseCond K F ν) (hN : N ≤ n) (hpair : 2*N ≤ n)
(w : Seed d u n × (Fin n → ℝ)) (v : Fin (d+1) → Space d)
(hseed : ∀ a k, ‖seedCandidates w.1 a k‖ ≤ (oracleConfig d D n).radius / (2*(d : ℝ)))
(hg : historyPrefix hpair w ∈ estimatorGood (oracleConfig d D n) K hK.isClosed.measurableSet R anchor ν hνm N v) :
∀ p < N, (interior (codedBody K (traceState (oracleConfig d D n) K R w.1 (terminalPairs w.2) p).1)).Nonempty → ∀ k,
|traceScalar (oracleConfig d D n) K R w.1 (terminalPairs w.2) p k -
traceGap (oracleConfig d D n) K R w.1 (terminalPairs w.2) F p k| ≤
confidenceRadius (oracleConfig d D n) u (traceSampleCount (oracleConfig d D n) K R w.1 (terminalPairs w.2) p) := by sorry