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

Lean module · EXP3

BanditRLProof.Exp3HighProbabilityRegret

# Generated predictable EXP3 high-probability regret This module completes the finite-horizon high-probability pseudo-regret route for the generated predictable EXP3 trajectory. A pathwise reciprocal-floor bound controls the random estimator-square sum almost surely. The final theorem then combines the sampled Hedge inequality, exploration bias, the pure-Hedge predictable-minus-observed confidence event, and the comparator-estimator confidence event by a two-event union bound. The resulting theorem is valid but uses range-based Hoeffding radii. Their `(|arms| / gamma)^2` per-round proxy is intentionally recorded rather than silently presented as the ideal EXP3.P/Freedman rate.

Module map

Declarations
6
Placeholders
0

Imports

BanditRLProof.Exp3PureConfidence

Imported by

BanditRLProof, BanditRLProof.Exp3BernsteinHighProbabilityRegret, BanditRLProof.Exp3PredictableRegretAllTime, BanditRLProof.Exp3RealizedHighProbabilityRegret

Declarations

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

theorem BanditRLProof.Exp3.observedMixedSquaredImportanceWeightedLossAt_le_inv_explorationFloor Compiled

No declaration docstring is present; use the chapter context and exact statement below.

theorem observedMixedSquaredImportanceWeightedLossAt_le_inv_explorationFloor {Env : Type u} {Action : Type v} [MeasurableSpace Env] [MeasurableSpace Action] [MeasurableSingletonClass Action] [DecidableEq Action] (arms : Finset Action) (harms : arms.Nonempty) (eta gamma : Real) (hgamma_pos : 0 < gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) (t : Nat) (sample : Env × ((k : Nat) -> Action × Real)) (hreward : (sample.2 t).2 ∈ Set.Icc (0 : Real) 1) : observedMixedSquaredImportanceWeightedLossAt arms eta gamma t sample <= 1 / (gamma / (arms.card : Real))
theorem BanditRLProof.Exp3.sampledPredictableTrajectoryMeasure_reward_mem_unitInterval_ae Compiled

No declaration docstring is present; use the chapter context and exact statement below.

theorem sampledPredictableTrajectoryMeasure_reward_mem_unitInterval_ae {Env : Type u} {Action : Type v} [MeasurableSpace Env] [StandardBorelSpace 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_nonneg : 0 <= gamma) (hgamma_le_one : gamma <= 1) (loss : PredictableLossVector Env Action) (t : Nat) : let mu := prior ⊗ₘ sampledImportanceWeightedTrajectoryKernel arms harms eta gamma hgamma_nonneg hgamma_le_one loss.environment ∀ᵐ sample ∂mu, (sample.2 t).2 ∈ Set.Icc (0 : Real) 1
theorem BanditRLProof.Exp3.sampledPredictableTrajectoryMeasure_observedMixedSquared_sum_le_ae Compiled

No declaration docstring is present; use the chapter context and exact statement below.

theorem sampledPredictableTrajectoryMeasure_observedMixedSquared_sum_le_ae {Env : Type u} {Action : Type v} [MeasurableSpace Env] [StandardBorelSpace 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) (horizon : Nat) : let mu := prior ⊗ₘ sampledImportanceWeightedTrajectoryKernel arms harms eta gamma hgamma_pos.le hgamma_le_one loss.environment ∀ᵐ sample ∂mu, (Finset.range horizon).sum (fun t => observedMixedSquaredImportanceWeightedLossAt arms eta gamma t sample) <= (horizon : Real) * (1 / (gamma / (arms.card : Real)))
def BanditRLProof.Exp3.sampledPredictableHighProbabilityRegretBudget Compiled

No declaration docstring is present; use the chapter context and exact statement below.

noncomputable def sampledPredictableHighProbabilityRegretBudget {Action : Type v} (arms : Finset Action) (eta gamma : Real) (horizon : Nat) (delta : Real) : Real
theorem BanditRLProof.Exp3.sampledPredictable_highProbabilityRegret_tail_delta Compiled

Generated predictable EXP3 pseudo-regret exceeds the explicit Hedge, exploration, estimator-square, and two confidence-radius budget with probability at most the sum of the pure-q and comparator failure probabilities.

theorem sampledPredictable_highProbabilityRegret_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) (heta : 0 < eta) (hgamma_pos : 0 < gamma) (hgamma_lt_one : gamma < 1) (loss : PredictableLossVector Env Action) (comparator : Action) (hcomparator : comparator ∈ arms) (horizon : Nat) (hhorizon : 0 < horizon) (delta : Real) (hdelta : 0 < delta) : let mu := prior ⊗ₘ sampledImportanceWeightedTrajectoryKernel arms harms eta gamma hgamma_pos.le hgamma_lt_one.le loss.environment mu {sample | sampledPredictableHighProbabilityRegretBudget arms eta gamma horizon delta <= (Finset.range horizon).sum (fun t => sampledTrajectoryExploredPredictableLossAt arms eta gamma loss t sample) - (Finset.range horizon).sum (fun t => predictableLossAt loss t sample comparator)} <= ENNReal.ofReal delta + ENNReal.ofReal delta
theorem BanditRLProof.Exp3.sampledPredictable_highProbabilityRegret_tail_total_delta Compiled

Standard total-failure-probability form of the generated predictable EXP3 pseudo-regret bound. Each of the two confidence events receives `delta / 2`, so their union has probability at most `delta`.

theorem sampledPredictable_highProbabilityRegret_tail_total_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) (heta : 0 < eta) (hgamma_pos : 0 < gamma) (hgamma_lt_one : gamma < 1) (loss : PredictableLossVector Env Action) (comparator : Action) (hcomparator : comparator ∈ arms) (horizon : Nat) (hhorizon : 0 < horizon) (delta : Real) (hdelta : 0 < delta) : let mu := prior ⊗ₘ sampledImportanceWeightedTrajectoryKernel arms harms eta gamma hgamma_pos.le hgamma_lt_one.le loss.environment mu {sample | sampledPredictableHighProbabilityRegretBudget arms eta gamma horizon (delta / 2) <= (Finset.range horizon).sum (fun t => sampledTrajectoryExploredPredictableLossAt arms eta gamma loss t sample) - (Finset.range horizon).sum (fun t => predictableLossAt loss t sample comparator)} <= ENNReal.ofReal delta