BanditRLlib
Lean gate passed before this site build; local proof declarations are shown as compiled.

Lean module · Foundations

BanditRLProof.ExpectationWeightedPullCount

# Weighted lower-integral pull-count identities This module proves the nonnegative weighted-count lower-integral bridge. It remains in `ENNReal`, with an arbitrary finite action set and arbitrary nonnegative gap weights, before any `Rat`/`Real` or Bochner-expectation route.

Module map

Declarations
1
Placeholders
0

Imports

BanditRLProof.ExpectationPullCount, BanditRLProof.MeasurablePullCountCast

Imported by

BanditRLProof, BanditRLProof.ExpectationWeightedPullCountBounds

Declarations

Open an item to read its exact compact statement and source link. Detailed teaching notes are linked when registered.

theorem BanditRLProof.lintegral_finset_sum_gap_mul_natCast_pullCount_eq Compiled

The lower integral of a finite weighted sum of scalar-casted pull counts equals the same finite weighted sum of the corresponding action-event measures. This is the `EXP-WEIGHTED-PULLCOUNT-LINTEGRAL` bridge. It is shaped like a nonnegative expected-regret identity, but it deliberately avoids `FiniteBanditModel`, `Rat`, `Real`, Bochner expectation, filtrations, kernels, and concentration assumptions.

theorem lintegral_finset_sum_gap_mul_natCast_pullCount_eq {Omega : Type u} {Action : Type v} [MeasurableSpace Omega] [MeasurableSpace Action] [MeasurableSingletonClass Action] [DecidableEq Action] (mu : Measure Omega) (action : Omega -> ActionTrace Action) (haction : forall t : Nat, Measurable (fun omega : Omega => action omega t)) (gap : Action -> ENNReal) (arms : Finset Action) (n : Nat) : MeasureTheory.lintegral mu (fun omega : Omega => arms.sum (fun a : Action => gap a * ((pullCount (action omega) a n : Nat) : ENNReal))) = arms.sum (fun a : Action => gap a * (Finset.range n).sum (fun t : Nat => mu {omega : Omega | action omega t = a}))