OAI/Combinatorics/EditApproximation/Refinement/PhysicalNodeRegret.lean
25 declarations, 0 contain sorry, 0 declared as axioms
First crossing for the literal physical refinement
Online and selected-group candidates are connected to the complete refinement by membership in its actual finite lists. Group success is the literal simultaneous estimator-error test, rather than a cone-bound assumption. The child environment remains fixed throughout this argument.
- theorem
OAI.EditApproximation.representativeBandActions_zero_connectionline 24 - theorem
OAI.EditApproximation.sharedBandRationalAction_wideline 37 - theorem
OAI.EditApproximation.rationalOnlineMinimum_le_memberline 51 - theorem
OAI.EditApproximation.rationalWideStep_online_existsline 61 - theorem
OAI.EditApproximation.rationalWideStepCone_le_groupline 78 - theorem
OAI.EditApproximation.rationalWideStepCone_le_wideline 88 - def
OAI.EditApproximation.unskippedGroupFailureline 100Source documentation
The actual per-group failure test uses the rational estimated costs.
def unskippedGroupFailure {α : Type*} (source target : List α) {M : ℕ} (parent : TargetInterval source.length) (b P thetaExponent : ℕ) (tau F : ℚ) (child initial value : Fin M → TargetInterval target.length → ℚ) (Q : ℕ) (members : List (UnskippedBand source target parent b P thetaExponent tau F child initial)) (draw : ∀ scale : Fin (groupDyadicScales M b F).length, Fin (groupSampleCount M Q b (groupDyadicScaleAt M b F scale)) → Fin M) (gamma : ℚ) : Bool := …Mentions: OAI.EditApproximation.TargetInterval, OAI.EditApproximation.UnskippedBand, OAI.EditApproximation.groupDyadicScales, OAI.EditApproximation.groupSampleCount, OAI.EditApproximation.groupDyadicScaleAt
- theorem
OAI.EditApproximation.unskippedGroupFailure_falseline 111 - theorem
OAI.EditApproximation.rationalWideStepCone_group_boundline 132Source documentation
A selected group candidate in the actual list supplies the cone bound on the literal estimator-success event.
theorem rationalWideStepCone_group_bound {α : Type*} (source target : List α) {M : ℕ} (parent : TargetInterval source.length) (representatives : List (ℕ × TargetInterval target.length)) (b P F thetaExponent Q : ℕ) (tau a : ℚ) (parentInitial : TargetInterval target.length → ℚ) (child initial : Fin M → TargetInterval target.length → ℚ) (old current : Fin M × TargetInterval target.length → ℚ) (cell : ℕ × ℕ) (hnonempty : cellUnskippedBands source target parent b P F thetaExponent tau a parentInitial child initial cell ≠ []) (draw : ∀ scale : Fin (groupDyadicScales M b F).length, Fin (groupSampleCount M Q b (groupDyadicScaleAt M b F scale)) → Fin M) (specified : UnskippedBand source target parent b P thetaExponent tau F child initial) (hspecified : specified ∈ cellUnskippedBands source target parent b P F thetaExponent tau a …Mentions: OAI.EditApproximation.TargetInterval, OAI.EditApproximation.cellUnskippedBands, OAI.EditApproximation.groupDyadicScales, OAI.EditApproximation.groupSampleCount, OAI.EditApproximation.groupDyadicScaleAt, OAI.EditApproximation.UnskippedBand
- def
OAI.EditApproximation.physicalNodeChildrenline 177 - def
OAI.EditApproximation.physicalNodeGroupsline 180 - def
OAI.EditApproximation.physicalNodeOnlineline 183 - def
OAI.EditApproximation.physicalNodeConeline 186 - def
OAI.EditApproximation.physicalNodeOnlineValueline 194 - theorem
OAI.EditApproximation.physicalNode_next_eqline 210Source documentation
Literal candidate membership makes the online option present.
theorem physicalNode_next_eq (hleaf : ¬ (parent).hi - (parent).lo ≤ 1) (t : ℕ) (q : TargetInterval target.length) (action : PhysicalRationalAction M target.length) (haction : action ∈ online t q) : (physicalRefinement source target U N P F A T delta choices node (t + 1) q : ℝ) = refinementUpdate (delta : ℝ) (cone t q : ℝ) (onlineValue t q : ℝ) := by …Mentions: OAI.EditApproximation.TargetInterval, OAI.EditApproximation.PhysicalRationalAction, OAI.EditApproximation.physicalRefinement, OAI.EditApproximation.refinementUpdate
- theorem
OAI.EditApproximation.physicalNode_online_leline 225Source documentation
The actual online minimum is at most any listed shared band action.
theorem physicalNode_online_le (t : ℕ) (q r : TargetInterval target.length) (action : Fin M → TargetInterval target.length) (haction : sharedBandRationalAction r action ∈ online t q) : (onlineValue t q : ℝ) ≤ (L : ℝ) * targetEndpointDistance q r + ∑ i, (history (t + 1) (i, action i) : ℝ) := by …Mentions: OAI.EditApproximation.TargetInterval, OAI.EditApproximation.sharedBandRationalAction, OAI.EditApproximation.targetEndpointDistance
- theorem
OAI.EditApproximation.physicalNode_cone_le_oldline 242Source documentation
The old-index wide candidate remains one of the literal cone options.
theorem physicalNode_cone_le_old (b : ℕ) (tau theta : ℚ) (r q : TargetInterval target.length) (child : Fin M → TargetInterval target.length → ℚ) (hrep : (b, r) ∈ representatives) (action : Fin M → TargetInterval target.length) (haction : action ∈ representativeBandActions source target parent r b P tau theta child) (t : ℕ) : (cone t q : ℝ) ≤ (L : ℝ) * targetEndpointDistance q r + ∑ i, (history t (i, action i) : ℝ) := by …Mentions: OAI.EditApproximation.TargetInterval, OAI.EditApproximation.representativeBandActions, OAI.EditApproximation.targetEndpointDistance
- theorem
OAI.EditApproximation.physicalNode_comparison_lowerline 257Source documentation
True cost is bounded by the actual connection charge and clipped comparison children; no assumption about clipped warmup masses is needed.
theorem physicalNode_comparison_lower (hM : 0 < M) (hF : 1 ≤ F) (b : ℕ) (tau theta : ℚ) (r q : TargetInterval target.length) (child : Fin M → TargetInterval target.length → ℚ) (comparison : Fin M → TargetInterval target.length) (hc : comparison ∈ representativeBandActions source target parent r b P tau theta child) (t : ℕ) : (stateEditCost (physicalSource source node) target q : ℝ) ≤ (L : ℝ) * targetEndpointDistance q r + ∑ i, max (stateEditCost (physicalSource source (physicalChild node hbelow i)) target (comparison i) : ℝ) (history t (i, comparison i) : ℝ) := by …Mentions: OAI.EditApproximation.TargetInterval, OAI.EditApproximation.representativeBandActions, OAI.EditApproximation.stateEditCost, OAI.EditApproximation.physicalSource, OAI.EditApproximation.targetEndpointDistance, OAI.EditApproximation.physicalChild
- def
OAI.EditApproximation.physicalComparisonClippedline 285 - def
OAI.EditApproximation.physicalNodeCellBandsline 290 - def
OAI.EditApproximation.physicalNodeSelectedGroupline 295 - def
OAI.EditApproximation.physicalNodeGroupFailureline 302 - theorem
OAI.EditApproximation.physicalNode_group_boundline 320Source documentation
The selected whole-cell representative supplies the actual physical cone inequality on the actual estimator-success event.
theorem physicalNode_group_bound (b thetaExponent Q : ℕ) (tau gamma : ℚ) (htau : 0 < tau) (hdelta : 0 ≤ delta) (cell : ℕ × ℕ) (hnonempty : physicalNodeCellBands source target U N P F A T node hbelow b thetaExponent tau cell ≠ []) (specified : UnskippedBand source target parent b P thetaExponent tau F (physicalBandChildMassTable source target U N P F A T node hbelow) (physicalBandChildInitial source target U N P F A node hbelow)) (hspecified : specified ∈ physicalNodeCellBands source target U N P F A T node hbelow b thetaExponent tau cell) (t : ℕ) (draw : ∀ scale : Fin (groupDyadicScales M b F).length, Fin (groupSampleCount M Q b (groupDyadicScaleAt M b F scale)) → Fin M) (hselected : (physicalNodeSelectedGroup source target U N P F A T delta choices node hbelow b thetaExponent Q tau cell hnonempty t draw).rationalAction ∈ groups t) (hfailure : physicalNodeGroupFailure source target U N P F A T delta choices node hbelow …Mentions: OAI.EditApproximation.physicalNodeCellBands, OAI.EditApproximation.UnskippedBand, OAI.EditApproximation.physicalBandChildMassTable, OAI.EditApproximation.physicalBandChildInitial, OAI.EditApproximation.groupDyadicScales, OAI.EditApproximation.groupSampleCount, OAI.EditApproximation.groupDyadicScaleAt, OAI.EditApproximation.physicalNodeSelectedGroup, OAI.EditApproximation.physicalNodeGroupFailure
- theorem
OAI.EditApproximation.physicalNode_clipped_upperline 362 - theorem
OAI.EditApproximation.physicalNode_firstCrossing_regretline 378Source documentation
Deterministic first-crossing transfer for the complete physical recurrence, using its actual sampled candidate lists and group-error test.
theorem physicalNode_firstCrossing_regret (hleaf : ¬ (parent).hi - (parent).lo ≤ 1) (b thetaExponent Q : ℕ) (tau gamma κ : ℚ) (htau : 0 < tau) (hdelta : 0 < delta) (hκ : 0 ≤ κ) (hκhalf : κ ≤ 1 / 2) (cell : ℕ × ℕ) (hnonempty : physicalNodeCellBands source target U N P F A T node hbelow b thetaExponent tau cell ≠ []) (specified : UnskippedBand source target parent b P thetaExponent tau F (physicalBandChildMassTable source target U N P F A T node hbelow) (physicalBandChildInitial source target U N P F A node hbelow)) (hspecified : specified ∈ physicalNodeCellBands source target U N P F A T node hbelow b thetaExponent tau cell) (hrep : (b, specified.center) ∈ representatives) (q : TargetInterval target.length) (hb : (b : ℝ) ≤ stateEditCost (physicalSource source node) target q) …Mentions: OAI.EditApproximation.physicalNodeCellBands, OAI.EditApproximation.UnskippedBand, OAI.EditApproximation.physicalBandChildMassTable, OAI.EditApproximation.physicalBandChildInitial, OAI.EditApproximation.TargetInterval, OAI.EditApproximation.stateEditCost, OAI.EditApproximation.physicalSource