Lean module · EXP3
BanditRLProof.Exp3PureConfidence
# Pure-Hedge cross-weighted EXP3 confidence This module proves finite-horizon confidence bounds for the pure-Hedge weighted importance estimator, including the predictable-minus-observed direction needed by the regret decomposition. It first identifies the conditional mean under the exploration action law, transports the latent predictable score to the observed trajectory score, proves adaptedness, and applies the local conditional sub-Gaussian finite-sum tail theorem. The bounded-range proxy is Hoeffding's interval proxy for `[0, K / gamma]`. This closes the pure-q cross-weight concentration leaf; it does not control the random estimator-square term in the sampled Hedge inequality.
Module map
Imports
BanditRLProof.Exp3ComparatorConfidence
Imported by
BanditRLProof, BanditRLProof.Exp3HighProbabilityRegret, BanditRLProof.Exp3PureBernstein
Declarations
Open an item to read its exact compact statement and source link. Detailed teaching notes are linked when registered.
theorem
BanditRLProof.Exp3.weightedEstimator_hasCondSubgaussianMGF_of_condDistrib_ae_eq_finiteActionKernel
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
theorem weightedEstimator_hasCondSubgaussianMGF_of_condDistrib_ae_eq_finiteActionKernel {Omega : Type u} {History : Type v} {Action : Type w} [mOmega : MeasurableSpace Omega] [StandardBorelSpace Omega] [Nonempty Omega] [mHistory : MeasurableSpace History] [StandardBorelSpace History] [mAction : MeasurableSpace Action] [MeasurableSingletonClass Action] [StandardBorelSpace Action] [Nonempty Action] [DecidableEq Action] (mu : Measure Omega) [IsFiniteMeasure mu] (history : Omega -> History) (hhistory : Measurable history) (action : Omega -> Action) (haction : Measurable action) (arms : Finset Action) (prob weight loss : History -> Action -> Real) (probSource : MeasurableFiniteActionDistribution arms prob) (weightSource : MeasurableFiniteActionDistribution arms weight) (epsilon : Real) (regularity : BoundedMeasurableLossWithProbabilityFloor arms prob loss epsilon) (hcond : condDistrib action history mu =ᵐ[mu.map history] finiteActionKernel arms prob probSource) : ProbabilityTheory.HasCondSubgaussianMGF (mHistory.comap history) hhistory.comap_le (fun omega => weightedImportanceWeightedLoss arms (prob (history omega)) (weight (history omega)) (loss (history omega)) (action omega) - arms.sum (fun candidate => weight (history omega) candidate * loss (history omega) candidate)) (Concentration.intervalVarianceProxy 0 (1 / epsilon)) mu
def
BanditRLProof.Exp3.sampledTrajectoryWeightedPurePredictableDeviationAt
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
noncomputable def sampledTrajectoryWeightedPurePredictableDeviationAt {Env : Type u} {Action : Type v} [MeasurableSpace Env] [MeasurableSpace Action] [DecidableEq Action] (arms : Finset Action) (eta gamma : Real) (loss : PredictableLossVector Env Action) (t : Nat) (sample : Env × ((k : Nat) -> Action × Real)) : Real
def
BanditRLProof.Exp3.sampledTrajectoryPureObservedDeviationAt
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
noncomputable def sampledTrajectoryPureObservedDeviationAt {Env : Type u} {Action : Type v} [MeasurableSpace Env] [MeasurableSpace Action] [DecidableEq Action] (arms : Finset Action) (eta gamma : Real) (loss : PredictableLossVector Env Action) (t : Nat) (sample : Env × ((k : Nat) -> Action × Real)) : Real
theorem
BanditRLProof.Exp3.sampledWeightedPurePredictableDeviation_zero_hasCondSubgaussianMGF
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
theorem sampledWeightedPurePredictableDeviation_zero_hasCondSubgaussianMGF {Env : Type u} {Action : Type v} [MeasurableSpace Env] [StandardBorelSpace Env] [Nonempty Env] [MeasurableSpace Action] [MeasurableSingletonClass Action] [StandardBorelSpace Action] [Nonempty Action] [DecidableEq Action] (prior : Measure Env) [IsFiniteMeasure prior] (arms : Finset Action) (harms : arms.Nonempty) (eta gamma : Real) (hgamma_pos : 0 < gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) : let mu := prior ⊗ₘ sampledImportanceWeightedTrajectoryKernel arms harms eta gamma hgamma_pos.le hgamma_le_one loss.environment ProbabilityTheory.HasCondSubgaussianMGF ((inferInstance : MeasurableSpace Env).comap (fun sample : Env × ((k : Nat) -> Action × Real) => sample.1)) measurable_fst.comap_le (sampledTrajectoryWeightedPurePredictableDeviationAt arms eta gamma loss 0) (sampledComparatorEstimatorVarianceProxy arms gamma) mu
theorem
BanditRLProof.Exp3.sampledWeightedPurePredictableDeviation_succ_hasCondSubgaussianMGF
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
theorem sampledWeightedPurePredictableDeviation_succ_hasCondSubgaussianMGF {Env : Type u} {Action : Type v} [MeasurableSpace Env] [StandardBorelSpace Env] [Nonempty Env] [MeasurableSpace Action] [MeasurableSingletonClass Action] [StandardBorelSpace Action] [Nonempty Action] [DecidableEq Action] (prior : Measure Env) [IsFiniteMeasure prior] (arms : Finset Action) (harms : arms.Nonempty) (eta gamma : Real) (hgamma_pos : 0 < gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) (n : Nat) : let mu := prior ⊗ₘ sampledImportanceWeightedTrajectoryKernel arms harms eta gamma hgamma_pos.le hgamma_le_one loss.environment let history := fun sample : Env × ((k : Nat) -> Action × Real) => (sample.1, Preorder.frestrictLe n sample.2) ProbabilityTheory.HasCondSubgaussianMGF ((inferInstance : MeasurableSpace (Env × History.FinitePairHistory Action Real n)).comap history) (measurable_fst.prodMk ((Preorder.measurable_frestrictLe n).comp measurable_snd)).comap_le (sampledTrajectoryWeightedPurePredictableDeviationAt arms eta gamma loss (n + 1)) (sampledComparatorEstimatorVarianceProxy arms gamma) mu
theorem
BanditRLProof.Exp3.sampledPureObservedDeviation_zero_hasCondSubgaussianMGF
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
theorem sampledPureObservedDeviation_zero_hasCondSubgaussianMGF {Env : Type u} {Action : Type v} [MeasurableSpace Env] [StandardBorelSpace Env] [Nonempty Env] [MeasurableSpace Action] [MeasurableSingletonClass Action] [StandardBorelSpace Action] [Nonempty Action] [DecidableEq Action] (prior : Measure Env) [IsFiniteMeasure prior] (arms : Finset Action) (harms : arms.Nonempty) (eta gamma : Real) (hgamma_pos : 0 < gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) : let mu := prior ⊗ₘ sampledImportanceWeightedTrajectoryKernel arms harms eta gamma hgamma_pos.le hgamma_le_one loss.environment ProbabilityTheory.HasCondSubgaussianMGF ((inferInstance : MeasurableSpace Env).comap (fun sample : Env × ((k : Nat) -> Action × Real) => sample.1)) measurable_fst.comap_le (sampledTrajectoryPureObservedDeviationAt arms eta gamma loss 0) (sampledComparatorEstimatorVarianceProxy arms gamma) mu
theorem
BanditRLProof.Exp3.sampledPureObservedDeviation_succ_hasCondSubgaussianMGF
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
theorem sampledPureObservedDeviation_succ_hasCondSubgaussianMGF {Env : Type u} {Action : Type v} [MeasurableSpace Env] [StandardBorelSpace Env] [Nonempty Env] [MeasurableSpace Action] [MeasurableSingletonClass Action] [StandardBorelSpace Action] [Nonempty Action] [DecidableEq Action] (prior : Measure Env) [IsFiniteMeasure prior] (arms : Finset Action) (harms : arms.Nonempty) (eta gamma : Real) (hgamma_pos : 0 < gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) (n : Nat) : let mu := prior ⊗ₘ sampledImportanceWeightedTrajectoryKernel arms harms eta gamma hgamma_pos.le hgamma_le_one loss.environment let history := fun sample : Env × ((k : Nat) -> Action × Real) => (sample.1, Preorder.frestrictLe n sample.2) ProbabilityTheory.HasCondSubgaussianMGF ((inferInstance : MeasurableSpace (Env × History.FinitePairHistory Action Real n)).comap history) (measurable_fst.prodMk ((Preorder.measurable_frestrictLe n).comp measurable_snd)).comap_le (sampledTrajectoryPureObservedDeviationAt arms eta gamma loss (n + 1)) (sampledComparatorEstimatorVarianceProxy arms gamma) mu
def
BanditRLProof.Exp3.sampledPureObservedDeviationProcess
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
noncomputable def sampledPureObservedDeviationProcess {Env : Type u} {Action : Type v} [MeasurableSpace Env] [MeasurableSpace Action] [DecidableEq Action] (arms : Finset Action) (eta gamma : Real) (loss : PredictableLossVector Env Action) : Nat -> Env × ((k : Nat) -> Action × Real) -> Real | 0, _sample => 0 | i + 1, sample => sampledTrajectoryPureObservedDeviationAt arms eta gamma loss i sample theorem sampledPureObservedDeviationProcess_stronglyAdapted {Env : Type u} {Action : Type v} [MeasurableSpace Env] [MeasurableSpace Action] [MeasurableSingletonClass Action] [DecidableEq Action] (arms : Finset Action) (harms : arms.Nonempty) (eta gamma : Real) (hgamma_nonneg : 0 <= gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) : StronglyAdapted (sampledPredictableDeviationFiltration Env Action) (sampledPureObservedDeviationProcess arms eta gamma loss)
theorem
BanditRLProof.Exp3.sampledPureObservedDeviationProcess_stronglyAdapted
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
theorem sampledPureObservedDeviationProcess_stronglyAdapted {Env : Type u} {Action : Type v} [MeasurableSpace Env] [MeasurableSpace Action] [MeasurableSingletonClass Action] [DecidableEq Action] (arms : Finset Action) (harms : arms.Nonempty) (eta gamma : Real) (hgamma_nonneg : 0 <= gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) : StronglyAdapted (sampledPredictableDeviationFiltration Env Action) (sampledPureObservedDeviationProcess arms eta gamma loss)
abbrev
BanditRLProof.Exp3.sampledPureObservedDeviationVarianceProxy
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
noncomputable abbrev sampledPureObservedDeviationVarianceProxy {Action : Type v} (arms : Finset Action) (gamma : Real) : NNReal
abbrev
BanditRLProof.Exp3.sampledPureObservedDeviationProxy
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
noncomputable abbrev sampledPureObservedDeviationProxy {Action : Type v} (arms : Finset Action) (gamma : Real) : Nat -> NNReal
theorem
BanditRLProof.Exp3.sampledPureObservedDeviationProcess_sum_range_succ
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
theorem sampledPureObservedDeviationProcess_sum_range_succ {Env : Type u} {Action : Type v} [MeasurableSpace Env] [MeasurableSpace Action] [DecidableEq Action] (arms : Finset Action) (eta gamma : Real) (loss : PredictableLossVector Env Action) (horizon : Nat) (sample : Env × ((k : Nat) -> Action × Real)) : (Finset.range (horizon + 1)).sum (fun i => sampledPureObservedDeviationProcess arms eta gamma loss i sample) = (Finset.range horizon).sum (fun i => sampledTrajectoryPureObservedDeviationAt arms eta gamma loss i sample)
theorem
BanditRLProof.Exp3.sampledPureObservedDeviation_sum_tail_ennreal
Compiled
One-sided concentration for the pure-Hedge cross-weighted observed EXP3 estimator minus its true pure-Hedge predictable loss.
theorem sampledPureObservedDeviation_sum_tail_ennreal {Env : Type u} {Action : Type v} [MeasurableSpace Env] [StandardBorelSpace Env] [Nonempty Env] [MeasurableSpace Action] [MeasurableSingletonClass Action] [StandardBorelSpace Action] [Nonempty Action] [DecidableEq Action] (prior : Measure Env) [IsProbabilityMeasure prior] (arms : Finset Action) (harms : arms.Nonempty) (eta gamma : Real) (hgamma_pos : 0 < gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) (horizon : Nat) {eps : Real} (heps : 0 <= eps) : let mu := prior ⊗ₘ sampledImportanceWeightedTrajectoryKernel arms harms eta gamma hgamma_pos.le hgamma_le_one loss.environment mu {sample | eps <= (Finset.range horizon).sum (fun i => sampledTrajectoryPureObservedDeviationAt arms eta gamma loss i sample)} <= ENNReal.ofReal (Real.exp (-eps ^ 2 / (2 * ((((horizon : NNReal) * sampledPureObservedDeviationVarianceProxy arms gamma : NNReal)) : Real))))
def
BanditRLProof.Exp3.sampledPureObservedDeviationConfidenceRadius
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
noncomputable def sampledPureObservedDeviationConfidenceRadius {Action : Type v} (arms : Finset Action) (gamma : Real) (horizon : Nat) (delta : Real) : Real
theorem
BanditRLProof.Exp3.sampledPureObservedDeviation_sum_tail_exp_neg_budget
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
theorem sampledPureObservedDeviation_sum_tail_exp_neg_budget {Env : Type u} {Action : Type v} [MeasurableSpace Env] [StandardBorelSpace Env] [Nonempty Env] [MeasurableSpace Action] [MeasurableSingletonClass Action] [StandardBorelSpace Action] [Nonempty Action] [DecidableEq Action] (prior : Measure Env) [IsProbabilityMeasure prior] (arms : Finset Action) (harms : arms.Nonempty) (eta gamma : Real) (hgamma_pos : 0 < gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) (horizon : Nat) (hhorizon : 0 < horizon) (budget : Real) : let mu := prior ⊗ₘ sampledImportanceWeightedTrajectoryKernel arms harms eta gamma hgamma_pos.le hgamma_le_one loss.environment mu {sample | Real.sqrt (2 * ((((horizon : NNReal) * sampledPureObservedDeviationVarianceProxy arms gamma : NNReal)) : Real) * budget) <= (Finset.range horizon).sum (fun i => sampledTrajectoryPureObservedDeviationAt arms eta gamma loss i sample)} <= ENNReal.ofReal (Real.exp (-budget))
theorem
BanditRLProof.Exp3.sampledPureObservedDeviation_sum_tail_delta
Compiled
Delta-shaped one-sided confidence bound for the pure-Hedge cross-weighted observed EXP3 estimator against its true pure-Hedge predictable loss.
theorem sampledPureObservedDeviation_sum_tail_delta {Env : Type u} {Action : Type v} [MeasurableSpace Env] [StandardBorelSpace Env] [Nonempty Env] [MeasurableSpace Action] [MeasurableSingletonClass Action] [StandardBorelSpace Action] [Nonempty Action] [DecidableEq Action] (prior : Measure Env) [IsProbabilityMeasure prior] (arms : Finset Action) (harms : arms.Nonempty) (eta gamma : Real) (hgamma_pos : 0 < gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) (horizon : Nat) (hhorizon : 0 < horizon) (delta : Real) (hdelta : 0 < delta) : let mu := prior ⊗ₘ sampledImportanceWeightedTrajectoryKernel arms harms eta gamma hgamma_pos.le hgamma_le_one loss.environment mu {sample | sampledPureObservedDeviationConfidenceRadius arms gamma horizon delta <= (Finset.range horizon).sum (fun i => sampledTrajectoryPureObservedLossAt arms eta gamma i sample - sampledTrajectoryPurePredictableLossAt arms eta gamma loss i sample)} <= ENNReal.ofReal delta
def
BanditRLProof.Exp3.sampledTrajectoryPurePredictableMinusObservedAt
Compiled
The pure-Hedge predictable loss minus its observed cross-weighted estimator. This is the sign needed when the sampled Hedge inequality is converted to true predictable regret.
noncomputable def sampledTrajectoryPurePredictableMinusObservedAt {Env : Type u} {Action : Type v} [MeasurableSpace Env] [MeasurableSpace Action] [DecidableEq Action] (arms : Finset Action) (eta gamma : Real) (loss : PredictableLossVector Env Action) (t : Nat) (sample : Env × ((k : Nat) -> Action × Real)) : Real
theorem
BanditRLProof.Exp3.sampledPurePredictableMinusObserved_zero_hasCondSubgaussianMGF
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
theorem sampledPurePredictableMinusObserved_zero_hasCondSubgaussianMGF {Env : Type u} {Action : Type v} [MeasurableSpace Env] [StandardBorelSpace Env] [Nonempty Env] [MeasurableSpace Action] [MeasurableSingletonClass Action] [StandardBorelSpace Action] [Nonempty Action] [DecidableEq Action] (prior : Measure Env) [IsFiniteMeasure prior] (arms : Finset Action) (harms : arms.Nonempty) (eta gamma : Real) (hgamma_pos : 0 < gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) : let mu := prior ⊗ₘ sampledImportanceWeightedTrajectoryKernel arms harms eta gamma hgamma_pos.le hgamma_le_one loss.environment ProbabilityTheory.HasCondSubgaussianMGF ((inferInstance : MeasurableSpace Env).comap (fun sample : Env × ((k : Nat) -> Action × Real) => sample.1)) measurable_fst.comap_le (sampledTrajectoryPurePredictableMinusObservedAt arms eta gamma loss 0) (sampledPureObservedDeviationVarianceProxy arms gamma) mu
theorem
BanditRLProof.Exp3.sampledPurePredictableMinusObserved_succ_hasCondSubgaussianMGF
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
theorem sampledPurePredictableMinusObserved_succ_hasCondSubgaussianMGF {Env : Type u} {Action : Type v} [MeasurableSpace Env] [StandardBorelSpace Env] [Nonempty Env] [MeasurableSpace Action] [MeasurableSingletonClass Action] [StandardBorelSpace Action] [Nonempty Action] [DecidableEq Action] (prior : Measure Env) [IsFiniteMeasure prior] (arms : Finset Action) (harms : arms.Nonempty) (eta gamma : Real) (hgamma_pos : 0 < gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) (n : Nat) : let mu := prior ⊗ₘ sampledImportanceWeightedTrajectoryKernel arms harms eta gamma hgamma_pos.le hgamma_le_one loss.environment let history := fun sample : Env × ((k : Nat) -> Action × Real) => (sample.1, Preorder.frestrictLe n sample.2) ProbabilityTheory.HasCondSubgaussianMGF ((inferInstance : MeasurableSpace (Env × History.FinitePairHistory Action Real n)).comap history) (measurable_fst.prodMk ((Preorder.measurable_frestrictLe n).comp measurable_snd)).comap_le (sampledTrajectoryPurePredictableMinusObservedAt arms eta gamma loss (n + 1)) (sampledPureObservedDeviationVarianceProxy arms gamma) mu
def
BanditRLProof.Exp3.sampledPurePredictableMinusObservedProcess
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
noncomputable def sampledPurePredictableMinusObservedProcess {Env : Type u} {Action : Type v} [MeasurableSpace Env] [MeasurableSpace Action] [DecidableEq Action] (arms : Finset Action) (eta gamma : Real) (loss : PredictableLossVector Env Action) : Nat -> Env × ((k : Nat) -> Action × Real) -> Real | 0, _sample => 0 | i + 1, sample => sampledTrajectoryPurePredictableMinusObservedAt arms eta gamma loss i sample theorem sampledPurePredictableMinusObservedProcess_stronglyAdapted {Env : Type u} {Action : Type v} [MeasurableSpace Env] [MeasurableSpace Action] [MeasurableSingletonClass Action] [DecidableEq Action] (arms : Finset Action) (harms : arms.Nonempty) (eta gamma : Real) (hgamma_nonneg : 0 <= gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) : StronglyAdapted (sampledPredictableDeviationFiltration Env Action) (sampledPurePredictableMinusObservedProcess arms eta gamma loss)
theorem
BanditRLProof.Exp3.sampledPurePredictableMinusObservedProcess_stronglyAdapted
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
theorem sampledPurePredictableMinusObservedProcess_stronglyAdapted {Env : Type u} {Action : Type v} [MeasurableSpace Env] [MeasurableSpace Action] [MeasurableSingletonClass Action] [DecidableEq Action] (arms : Finset Action) (harms : arms.Nonempty) (eta gamma : Real) (hgamma_nonneg : 0 <= gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) : StronglyAdapted (sampledPredictableDeviationFiltration Env Action) (sampledPurePredictableMinusObservedProcess arms eta gamma loss)
theorem
BanditRLProof.Exp3.sampledPurePredictableMinusObservedProcess_sum_range_succ
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
theorem sampledPurePredictableMinusObservedProcess_sum_range_succ {Env : Type u} {Action : Type v} [MeasurableSpace Env] [MeasurableSpace Action] [DecidableEq Action] (arms : Finset Action) (eta gamma : Real) (loss : PredictableLossVector Env Action) (horizon : Nat) (sample : Env × ((k : Nat) -> Action × Real)) : (Finset.range (horizon + 1)).sum (fun i => sampledPurePredictableMinusObservedProcess arms eta gamma loss i sample) = (Finset.range horizon).sum (fun i => sampledTrajectoryPurePredictableMinusObservedAt arms eta gamma loss i sample)
theorem
BanditRLProof.Exp3.sampledPurePredictableMinusObserved_sum_tail_ennreal
Compiled
One-sided concentration in the sign required by the predictable-regret decomposition: true pure-Hedge predictable loss minus its observed estimator.
theorem sampledPurePredictableMinusObserved_sum_tail_ennreal {Env : Type u} {Action : Type v} [MeasurableSpace Env] [StandardBorelSpace Env] [Nonempty Env] [MeasurableSpace Action] [MeasurableSingletonClass Action] [StandardBorelSpace Action] [Nonempty Action] [DecidableEq Action] (prior : Measure Env) [IsProbabilityMeasure prior] (arms : Finset Action) (harms : arms.Nonempty) (eta gamma : Real) (hgamma_pos : 0 < gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) (horizon : Nat) {eps : Real} (heps : 0 <= eps) : let mu := prior ⊗ₘ sampledImportanceWeightedTrajectoryKernel arms harms eta gamma hgamma_pos.le hgamma_le_one loss.environment mu {sample | eps <= (Finset.range horizon).sum (fun i => sampledTrajectoryPurePredictableMinusObservedAt arms eta gamma loss i sample)} <= ENNReal.ofReal (Real.exp (-eps ^ 2 / (2 * ((((horizon : NNReal) * sampledPureObservedDeviationVarianceProxy arms gamma : NNReal)) : Real))))
theorem
BanditRLProof.Exp3.sampledPurePredictableMinusObserved_sum_tail_exp_neg_budget
Compiled
No declaration docstring is present; use the chapter context and exact statement below.
theorem sampledPurePredictableMinusObserved_sum_tail_exp_neg_budget {Env : Type u} {Action : Type v} [MeasurableSpace Env] [StandardBorelSpace Env] [Nonempty Env] [MeasurableSpace Action] [MeasurableSingletonClass Action] [StandardBorelSpace Action] [Nonempty Action] [DecidableEq Action] (prior : Measure Env) [IsProbabilityMeasure prior] (arms : Finset Action) (harms : arms.Nonempty) (eta gamma : Real) (hgamma_pos : 0 < gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) (horizon : Nat) (hhorizon : 0 < horizon) (budget : Real) : let mu := prior ⊗ₘ sampledImportanceWeightedTrajectoryKernel arms harms eta gamma hgamma_pos.le hgamma_le_one loss.environment mu {sample | Real.sqrt (2 * ((((horizon : NNReal) * sampledPureObservedDeviationVarianceProxy arms gamma : NNReal)) : Real) * budget) <= (Finset.range horizon).sum (fun i => sampledTrajectoryPurePredictableMinusObservedAt arms eta gamma loss i sample)} <= ENNReal.ofReal (Real.exp (-budget))
theorem
BanditRLProof.Exp3.sampledPurePredictableMinusObserved_sum_tail_delta
Compiled
Delta-shaped lower-tail counterpart in the sign needed by the generated predictable EXP3 regret decomposition.
theorem sampledPurePredictableMinusObserved_sum_tail_delta {Env : Type u} {Action : Type v} [MeasurableSpace Env] [StandardBorelSpace Env] [Nonempty Env] [MeasurableSpace Action] [MeasurableSingletonClass Action] [StandardBorelSpace Action] [Nonempty Action] [DecidableEq Action] (prior : Measure Env) [IsProbabilityMeasure prior] (arms : Finset Action) (harms : arms.Nonempty) (eta gamma : Real) (hgamma_pos : 0 < gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) (horizon : Nat) (hhorizon : 0 < horizon) (delta : Real) (hdelta : 0 < delta) : let mu := prior ⊗ₘ sampledImportanceWeightedTrajectoryKernel arms harms eta gamma hgamma_pos.le hgamma_le_one loss.environment mu {sample | sampledPureObservedDeviationConfidenceRadius arms gamma horizon delta <= (Finset.range horizon).sum (fun i => sampledTrajectoryPurePredictableLossAt arms eta gamma loss i sample - sampledTrajectoryPureObservedLossAt arms eta gamma i sample)} <= ENNReal.ofReal delta