Definition 3.1 — coin-bias estimators
DefinitionListReplicableBias_LowerBound_Supplement_Estimationlist-replicabilityp2o-batch-stoc27sp2o-gran-per-chapterp2o-plan-paperp2o-v1probabilitysample-complexity
Let , , and let be an -toss sampling algorithm on coins (the published model of §3.2). solves the -CoinBiasEstimationProblem if for every bias vector its output satisfies
The probability is over the coin tosses and the algorithm's internal randomness.
This is the accuracy notion behind the classical estimation lower bound (Lemma 3.2), which the -list lower bound uses for very large .
Formalization Note The failure output (none) never counts as an accurate estimate. Probabilities are outputProb, the published transcript-law probability of an output event.
Definition code
import Mathlib
import Definitions.Def_ListReplicableBias_LowerBound_Geometry
import Definitions.Def_ListReplicableBias_LowerBound_Sampling
noncomputable section
open MeasureTheory
open scoped ENNReal
namespace ListReplicableBias.LowerBound.Supplement
open ListReplicableBias.LowerBound
/-- Definition 3.1, PDF p. 3: an algorithm solves the (d, ε, δ)-coin-bias
estimation problem if for every bias vector in the unit cube its output lies
within sup-norm distance ε of the bias vector with probability at least 1 − δ.
The failure output `none` counts as a miss. -/
def IsBiasEstimator {d N : ℕ} (A : SamplingAlgorithm d N) (ε δ : ℝ) : Prop :=
∀ p ∈ cube d 0 1,
1 - δ ≤ outputProb A p (some '' {y : Fin d → ℝ | ‖y - p‖ ≤ ε})
end ListReplicableBias.LowerBound.Supplement
Source
Chen, Bastan, Pavan, Vinodchandran, The Price of List Replicability: Tight Bounds for Bias Estimation, STOC 2027 submission #681 (October 2026), https://github.com/sigact-stoc/stoc2027/releases/download/papers-1/681.pdf, Definition 3.1, PDF p. 3