OAI/Probability/MatroidSecretary/Sampling/FiniteSeedTheorems.lean

GitHub source file

16 declarations, 0 contain sorry, 0 declared as axioms

Exact finite-seed encoding, measurability, and law preservation.

  1. lemma OAI.MatroidProphet.finiteSeedEncode_injective line 15
  2. lemma OAI.MatroidProphet.finiteSeedDecode_encode line 27
  3. lemma OAI.MatroidProphet.measurable_finiteSeedEncode line 34
  4. lemma OAI.MatroidProphet.measurable_finiteSeedDecode line 38
  5. theorem OAI.MatroidProphet.finiteSeed_map_roundtrip line 42

    Source documentation

    The full original law is recovered, not merely the law of a used statistic.

    theorem finiteSeed_map_roundtrip (μ : Measure α) :
        (μ.map (finiteSeedEncode α)).map (finiteSeedDecode α) = μ := by
    …

    Mentions: OAI.MatroidProphet.finiteSeedEncode, OAI.MatroidProphet.finiteSeedDecode

  6. theorem OAI.MatroidProphet.finiteSeed_decode_measurePreserving line 49

    Source documentation

    Decoding the encoded law preserves every event of the original seed.

    theorem finiteSeed_decode_measurePreserving (μ : Measure α) :
        MeasurePreserving (finiteSeedDecode α) (μ.map (finiteSeedEncode α)) μ :=
    …

    Mentions: OAI.MatroidProphet.finiteSeedDecode, OAI.MatroidProphet.finiteSeedEncode

  7. theorem OAI.MatroidProphet.finiteSeed_integral_decode line 54

    Source documentation

    Arbitrary real payoff statistics have exactly the same expectation.

    theorem finiteSeed_integral_decode (μ : Measure α) (g : α → ℝ) :
        (∫ r, g (finiteSeedDecode α r) ∂μ.map (finiteSeedEncode α)) = ∫ a, g a ∂μ := by
    …

    Mentions: OAI.MatroidProphet.finiteSeedDecode, OAI.MatroidProphet.finiteSeedEncode

  8. theorem OAI.MatroidProphet.finiteSeed_indepFun_iff line 62

    Source documentation

    Encoding a full seed does not change its independence from any other data.

    theorem finiteSeed_indepFun_iff {Ω β : Type*} [MeasurableSpace Ω] [MeasurableSpace β]
        (μ : Measure Ω) (X : Ω → β) (R : Ω → α) :
        ProbabilityTheory.IndepFun X (finiteSeedEncode α ∘ R) μ ↔
          ProbabilityTheory.IndepFun X R μ := by
    …

    Mentions: OAI.MatroidProphet.finiteSeedEncode

  9. theorem OAI.MatroidProphet.reindexOnlineSeed_decisionAt line 78
  10. theorem OAI.MatroidProphet.reindexOnlineSeed_acceptedThrough line 82
  11. theorem OAI.MatroidProphet.reindexOnlineSeed_reward line 87
  12. theorem OAI.MatroidProphet.reindexOnlineSeed_feasible line 92

    Source documentation

    Total decoding also preserves feasibility on encodings outside the support.

    theorem reindexOnlineSeed_feasible (M : Matroid (Fin n)) (hA : Feasible M A) :
        Feasible M (reindexOnlineSeed A f) := by
    …

    Mentions: OAI.MatroidProphet.Feasible, OAI.MatroidProphet.reindexOnlineSeed

  13. theorem OAI.MatroidProphet.reindexHiddenSeed_acceptedThrough line 99
  14. theorem OAI.MatroidProphet.reindexHiddenSeed_reward line 104
  15. theorem OAI.MatroidProphet.reindexHiddenSeed_worstReward line 108
  16. theorem OAI.MatroidProphet.reindexHiddenSeed_integral line 112

    Source documentation

    No independence of the adversary from the reindexed seed is used.

    theorem reindexHiddenSeed_integral (ν' : Measure (Seed bits'))
        (ν : Measure (Seed bits)) (hlaw : ν'.map f = ν) (w : Weights n) :
        (∫ r, hiddenWorstReward (reindexHiddenSeed H f) w r ∂ν') =
          ∫ r, hiddenWorstReward H w r ∂ν := by
    …

    Mentions: OAI.MatroidProphet.Seed, OAI.MatroidProphet.Weights, OAI.MatroidProphet.hiddenWorstReward, OAI.MatroidProphet.reindexHiddenSeed