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

Lean module · OFUL

BanditRLProof.OFULUniformTimeConfidence

# Finite-window uniform OFUL confidence This module assembles the compiled deterministic-horizon scalar-ridge confidence theorem over every horizon in a finite inclusive window. It supports arbitrary confidence schedules with values in `(0, 1]` and an equal allocation whose budgets sum exactly to `delta`, giving failure probability at most `delta`.

Module map

Declarations
7
Placeholders
0

Imports

BanditRLProof.OFULScalarRegularizationBias, BanditRLProof.ProbabilityUnionBound

Imported by

BanditRLProof, BanditRLProof.OFULAllTimeConfidence, BanditRLProof.OFULGeneratedTrajectoryConfidenceGap

Declarations

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

def BanditRLProof.OFUL.scalarRidgeConfidenceFailureAt Compiled

Failure of the scalar-ridge confidence ellipsoid at one horizon.

def scalarRidgeConfidenceFailureAt {Omega : Type u} {Feature : Type v} [Fintype Feature] [DecidableEq Feature] (lambda : Real) (thetaStar : Feature -> Real) (S : Real) (feature : Nat -> Omega -> Feature -> Real) (response : Nat -> Omega -> Real) (R delta : Real) (n : Nat) : Set Omega
def BanditRLProof.OFUL.finiteHorizonScheduledScalarRidgeConfidenceFailureSet Compiled

Union of scalar-ridge confidence failures over all `n <= horizon`, with the confidence level selected by `deltaAt n`.

def finiteHorizonScheduledScalarRidgeConfidenceFailureSet {Omega : Type u} {Feature : Type v} [Fintype Feature] [DecidableEq Feature] (lambda : Real) (thetaStar : Feature -> Real) (S : Real) (feature : Nat -> Omega -> Feature -> Real) (response : Nat -> Omega -> Real) (R : Real) (deltaAt : Nat -> Real) (horizon : Nat) : Set Omega
theorem BanditRLProof.OFUL.mem_finiteHorizonScheduledScalarRidgeConfidenceFailureSet_iff Compiled

Membership in the scheduled failure set is failure at some `n <= horizon`.

theorem mem_finiteHorizonScheduledScalarRidgeConfidenceFailureSet_iff {Omega : Type u} {Feature : Type v} [Fintype Feature] [DecidableEq Feature] (lambda : Real) (thetaStar : Feature -> Real) (S : Real) (feature : Nat -> Omega -> Feature -> Real) (response : Nat -> Omega -> Real) (R : Real) (deltaAt : Nat -> Real) (horizon : Nat) (omega : Omega) : omega ∈ finiteHorizonScheduledScalarRidgeConfidenceFailureSet lambda thetaStar S feature response R deltaAt horizon ↔ ∃ n, n <= horizon ∧ omega ∈ scalarRidgeConfidenceFailureAt lambda thetaStar S feature response R (deltaAt n) n
theorem BanditRLProof.OFUL.not_mem_finiteHorizonScheduledScalarRidgeConfidenceFailureSet_iff Compiled

Outside the scheduled failure union, every scalar-ridge confidence ellipsoid in the inclusive finite window holds simultaneously.

theorem not_mem_finiteHorizonScheduledScalarRidgeConfidenceFailureSet_iff {Omega : Type u} {Feature : Type v} [Fintype Feature] [DecidableEq Feature] (lambda : Real) (thetaStar : Feature -> Real) (S : Real) (feature : Nat -> Omega -> Feature -> Real) (response : Nat -> Omega -> Real) (R : Real) (deltaAt : Nat -> Real) (horizon : Nat) (omega : Omega) : omega ∉ finiteHorizonScheduledScalarRidgeConfidenceFailureSet lambda thetaStar S feature response R deltaAt horizon ↔ ∀ n, n <= horizon -> matrixNorm (Matrix.scalar Feature lambda + finiteHorizonFeatureGram feature n omega) (finiteHorizonRidgeEstimate (Matrix.scalar Feature lambda) feature response n omega - thetaStar) <= finiteHorizonScalarConfidenceRadius feature R (deltaAt n) lambda S n omega
theorem BanditRLProof.OFUL.measure_finiteHorizonScheduledScalarRidgeConfidenceFailureSet_le_sum Compiled

Scheduled finite-window confidence: the failure-union probability is bounded by the sum of the fixed-time confidence budgets.

theorem measure_finiteHorizonScheduledScalarRidgeConfidenceFailureSet_le_sum {Omega : Type u} {Feature : Type v} [mOmega : MeasurableSpace Omega] [StandardBorelSpace Omega] [Fintype Feature] [DecidableEq Feature] [Nonempty Feature] (mu : Measure Omega) [IsProbabilityMeasure mu] (lambda : Real) (hlambda : 0 < lambda) (thetaStar : Feature -> Real) (S : Real) (htheta : euclideanLength thetaStar <= S) (F : Filtration Nat mOmega) (feature : Nat -> Omega -> Feature -> Real) (response noise : Nat -> Omega -> Real) (R : Real) (hR : 0 < R) (projectionBound : EuclideanSpace Real Feature -> Nat -> Real) (hfeature : forall i j, StronglyMeasurable[F i] (fun omega => feature i omega j)) (hnoise : StronglyAdapted F (fun t omega => match t with | 0 => 0 | i + 1 => noise i omega)) (hprojectionBound_nonneg : forall theta i, 0 <= projectionBound theta i) (hprojectionBound : forall theta i omega, |dotProduct (WithLp.ofLp theta) (feature i omega)| <= projectionBound theta i) (horizon : Nat) (hsubGaussian : forall i, i < horizon -> HasCondSubgaussianMGF (F i) (F.le i) (noise i) (constantSquaredVarianceProxy R i) mu) (hresponse : forall omega i, i < horizon -> response i omega = dotProduct thetaStar (feature i omega) + noise i omega) (deltaAt : Nat -> Real) (hdeltaAt : forall n, n <= horizon -> 0 < deltaAt n) (hdeltaAt_one : forall n, n <= horizon -> deltaAt n <= 1) : mu (finiteHorizonScheduledScalarRidgeConfidenceFailureSet lambda thetaStar S feature response R deltaAt horizon) <= (Finset.range (horizon + 1)).sum (fun n => ENNReal.ofReal (deltaAt n))
def BanditRLProof.OFUL.finiteHorizonUniformScalarRidgeConfidenceFailureSet Compiled

Equal-share finite-window failure union. The inclusive window has `horizon + 1` horizons, so each fixed-time theorem receives `delta / (horizon + 1)`.

def finiteHorizonUniformScalarRidgeConfidenceFailureSet {Omega : Type u} {Feature : Type v} [Fintype Feature] [DecidableEq Feature] (lambda : Real) (thetaStar : Feature -> Real) (S : Real) (feature : Nat -> Omega -> Feature -> Real) (response : Nat -> Omega -> Real) (R delta : Real) (horizon : Nat) : Set Omega
theorem BanditRLProof.OFUL.measure_finiteHorizonUniformScalarRidgeConfidenceFailureSet_le Compiled

Uniform scalar-ridge confidence over every deterministic horizon `n <= horizon`, obtained by equal allocation of the total failure budget.

theorem measure_finiteHorizonUniformScalarRidgeConfidenceFailureSet_le {Omega : Type u} {Feature : Type v} [mOmega : MeasurableSpace Omega] [StandardBorelSpace Omega] [Fintype Feature] [DecidableEq Feature] [Nonempty Feature] (mu : Measure Omega) [IsProbabilityMeasure mu] (lambda : Real) (hlambda : 0 < lambda) (thetaStar : Feature -> Real) (S : Real) (htheta : euclideanLength thetaStar <= S) (F : Filtration Nat mOmega) (feature : Nat -> Omega -> Feature -> Real) (response noise : Nat -> Omega -> Real) (R : Real) (hR : 0 < R) (projectionBound : EuclideanSpace Real Feature -> Nat -> Real) (hfeature : forall i j, StronglyMeasurable[F i] (fun omega => feature i omega j)) (hnoise : StronglyAdapted F (fun t omega => match t with | 0 => 0 | i + 1 => noise i omega)) (hprojectionBound_nonneg : forall theta i, 0 <= projectionBound theta i) (hprojectionBound : forall theta i omega, |dotProduct (WithLp.ofLp theta) (feature i omega)| <= projectionBound theta i) (horizon : Nat) (hsubGaussian : forall i, i < horizon -> HasCondSubgaussianMGF (F i) (F.le i) (noise i) (constantSquaredVarianceProxy R i) mu) (hresponse : forall omega i, i < horizon -> response i omega = dotProduct thetaStar (feature i omega) + noise i omega) (delta : Real) (hdelta : 0 < delta) (hdelta_one : delta <= 1) : mu (finiteHorizonUniformScalarRidgeConfidenceFailureSet lambda thetaStar S feature response R delta horizon) <= ENNReal.ofReal delta