§10.6, p. 38 — total variation metrizes weak convergence on finite support
ProvedRobustSAA.Discrete.tv_weakfinite-supportgoodness-of-fitp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1uniform-consistency
Let be a distribution on a known finite support of labels, and let be any sequence of distributions on that support. Write and for their probability vectors. Then
This identifies the topology used in uniform consistency with a concrete finite sum and allows the statistic bounds to control weak convergence.
Formalization Note Weak convergence is the topology on ProbabilityMeasure (Fin n), not a separately defined vector topology.
Preamble
import Mathlib import Definitions.Def_RobustSAA_Discrete_Setting
Formal statement
namespace RobustSAA.Discrete
open Filter MeasureTheory
theorem tv_weak (n : ℕ) (G : ℕ → ProbabilityMeasure (Fin n))
(F : ProbabilityMeasure (Fin n)) :
Tendsto G atTop (nhds F) ↔
Tendsto (fun N => dTV (pvec (G N)) (pvec F)) atTop (nhds 0) := by sorry
end RobustSAA.Discrete
Source
Bertsimas, Gupta, Kallus, Robust Sample Average Approximation, arXiv:1408.4445v3, §10.6, p. 38, first display
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.