OAI/Probability/MatroidSecretary/Labels/RelabelingTheorems.lean
18 declarations, 0 contain sorry, 0 declared as axioms
Exact transport from arbitrary finite labels to the Fin n information model.
- lemma
OAI.MatroidProphet.Relabeling.sum_le_finiteOptimumline 15 - lemma
OAI.MatroidProphet.Relabeling.finiteOptimum_nonnegline 24 - lemma
OAI.MatroidProphet.Relabeling.finiteOptimum_leline 28 - theorem
OAI.MatroidProphet.Relabeling.finiteOptimum_mapEquivline 39Source documentation
Changing finite labels preserves the exact offline optimum, for arbitrary real weights.
theorem finiteOptimum_mapEquiv [Fintype α] [Fintype β] (M : Matroid α) (e : α ≃ β) (w : α → ℝ) : finiteOptimum (M.mapEquiv e) (encodeWeights e w) = finiteOptimum M w := by …Mentions: OAI.MatroidProphet.Relabeling.finiteOptimum, OAI.MatroidProphet.Relabeling.encodeWeights
- theorem
OAI.MatroidProphet.Relabeling.finiteOptimum_finline 56 - theorem
OAI.MatroidProphet.Relabeling.encodeWeights_nonnegative_iffline 59 - theorem
OAI.MatroidProphet.Relabeling.mapEquiv_ground_univline 67 - theorem
OAI.MatroidProphet.Relabeling.mapEquiv_loop_iffline 71 - theorem
OAI.MatroidProphet.Relabeling.exists_fin_representationline 77Source documentation
Every finite labeled matroid has a full-ground
Fin nrepresentative with exactly the original offline benchmark and all loop labels retained.theorem exists_fin_representation [Fintype α] (M : Matroid α) (hE : M.E = Set.univ) : ∃ (e : α ≃ Fin (Fintype.card α)), (M.mapEquiv e).E = Set.univ ∧ (∀ w : α → ℝ, optimum (M.mapEquiv e) (encodeWeights e w) = finiteOptimum M w) ∧ (∀ a, (M.mapEquiv e).IsLoop (e a) ↔ M.IsLoop a) := by …Mentions: OAI.MatroidProphet.optimum, OAI.MatroidProphet.Relabeling.encodeWeights, OAI.MatroidProphet.Relabeling.finiteOptimum
- theorem
OAI.MatroidProphet.Relabeling.measurable_encodeWeightsline 89Source documentation
Coordinate reindexing is Borel, with no regularity assumption on weights.
theorem measurable_encodeWeights (e : α ≃ β) : Measurable (encodeWeights e) := by …
- theorem
OAI.MatroidProphet.Relabeling.measurable_encodeHistoryline 94 - theorem
OAI.MatroidProphet.Relabeling.measurable_transportedDecisionline 103Source documentation
The transported rule is measurable in precisely its permissible inputs.
theorem measurable_transportedDecision [Fintype α] [MeasurableSpace α] [MeasurableSingletonClass α] {n bits : ℕ} (e : α ≃ Fin n) (A : OnlineRule n bits) (k : Fin n) : Measurable (fun x : Seed bits × ((α → ℝ) × LabeledHistory α k) => transportedDecision e A k x.1 x.2.1 x.2.2) := by …Mentions: OAI.MatroidProphet.OnlineRule, OAI.MatroidProphet.Seed, OAI.MatroidProphet.Relabeling.LabeledHistory, OAI.MatroidProphet.Relabeling.transportedDecision
- theorem
OAI.MatroidProphet.Relabeling.encodeHistory_labeledHistoryline 113 - theorem
OAI.MatroidProphet.Relabeling.transportedDecision_eq_decisionAtline 120 - theorem
OAI.MatroidProphet.Relabeling.transportedAcceptedThrough_eq_mapline 128Source documentation
Exact accepted-set equality holds at every prefix, not just at termination.
theorem transportedAcceptedThrough_eq_map [Fintype α] {n bits : ℕ} (e : α ≃ Fin n) (A : OnlineRule n bits) (r : Seed bits) (s v : α → ℝ) (π : LabeledOrder α n) (t : ℕ) : transportedAcceptedThrough e A r s v π t = (acceptedThrough A r (encodeWeights e s) (encodeWeights e v) (π.trans e) t).map e.symm.toEmbedding := by …Mentions: OAI.MatroidProphet.OnlineRule, OAI.MatroidProphet.Seed, OAI.MatroidProphet.Relabeling.LabeledOrder, OAI.MatroidProphet.Relabeling.transportedAcceptedThrough, OAI.MatroidProphet.acceptedThrough, OAI.MatroidProphet.Relabeling.encodeWeights
- theorem
OAI.MatroidProphet.Relabeling.transportedReward_eq_rewardline 143Source documentation
The cardinal relabeling preserves the actual sum of accepted weights exactly.
theorem transportedReward_eq_reward [Fintype α] {n bits : ℕ} (e : α ≃ Fin n) (A : OnlineRule n bits) (r : Seed bits) (s v : α → ℝ) (π : LabeledOrder α n) : transportedReward e A r s v π = reward A r (encodeWeights e s) (encodeWeights e v) (π.trans e) := by …Mentions: OAI.MatroidProphet.OnlineRule, OAI.MatroidProphet.Seed, OAI.MatroidProphet.Relabeling.LabeledOrder, OAI.MatroidProphet.Relabeling.transportedReward, OAI.MatroidProphet.reward, OAI.MatroidProphet.Relabeling.encodeWeights
- theorem
OAI.MatroidProphet.Relabeling.transported_feasibleline 151Source documentation
Every-prefix feasibility transfers without dropping loops or exceptional inputs.
theorem transported_feasible [Fintype α] {n bits : ℕ} (M : Matroid α) (e : α ≃ Fin n) (A : OnlineRule n bits) (hA : Feasible (M.mapEquiv e) A) (r : Seed bits) (s v : α → ℝ) (π : LabeledOrder α n) (t : ℕ) (hs : ∀ a, 0 ≤ s a) (hv : ∀ a, 0 ≤ v a) : M.Indep (transportedAcceptedThrough e A r s v π t : Set α) := by …Mentions: OAI.MatroidProphet.OnlineRule, OAI.MatroidProphet.Feasible, OAI.MatroidProphet.Seed, OAI.MatroidProphet.Relabeling.LabeledOrder, OAI.MatroidProphet.Relabeling.transportedAcceptedThrough
- theorem
OAI.MatroidProphet.Relabeling.transportedAcceptedThrough_monoline 163Source documentation
The relabeled implementation retains every previous acceptance irrevocably.
theorem transportedAcceptedThrough_mono [Fintype α] {n bits : ℕ} (e : α ≃ Fin n) (A : OnlineRule n bits) (r : Seed bits) (s v : α → ℝ) (π : LabeledOrder α n) {t t' : ℕ} (htt' : t ≤ t') : transportedAcceptedThrough e A r s v π t ⊆ transportedAcceptedThrough e A r s v π t' := by …Mentions: OAI.MatroidProphet.OnlineRule, OAI.MatroidProphet.Seed, OAI.MatroidProphet.Relabeling.LabeledOrder, OAI.MatroidProphet.Relabeling.transportedAcceptedThrough