math / data /lean /OAI /Algorithms /SubexponentialSAT /SqrtLog /BucketProgramProducedRankMatrix.lean
Download data/lean/OAI/Algorithms/SubexponentialSAT/SqrtLog/BucketProgramProducedRankMatrix.lean from introvoyz045/math: direct link, hf CLI and curl.
- Browser
- Download file 4.73 kB
-
https://huggingface.co/datasets/introvoyz045/math/resolve/main/data/lean/OAI/Algorithms/SubexponentialSAT/SqrtLog/BucketProgramProducedRankMatrix.lean
- Command line
-
hf download hf://datasets/introvoyz045/math/data/lean/OAI/Algorithms/SubexponentialSAT/SqrtLog/BucketProgramProducedRankMatrix.lean
-
curl -L -o BucketProgramProducedRankMatrix.lean https://huggingface.co/datasets/introvoyz045/math/resolve/main/data/lean/OAI/Algorithms/SubexponentialSAT/SqrtLog/BucketProgramProducedRankMatrix.lean
4.73 kB
| import OAI.Algorithms.SubexponentialSAT.LeanCompat | |
| import OAI.Algorithms.SubexponentialSAT.SqrtLog.BucketProgramRank | |
| import OAI.Algorithms.SubexponentialSAT.SqrtLog.LayoutMatrixProgramCompatibility | |
| import OAI.Algorithms.SubexponentialSAT.SqrtLog.BucketTupleProgramSource | |
| namespace OAI.Algorithms | |
| /-! The exact row-major matrix emitted by the entry scanner, in the actual | |
| selected identifier order. These proofs introduce no runtime permutation. -/ | |
| namespace SubexponentialSAT.SqrtLog.BucketProgram.ProducedRank | |
| private theorem forall₂_typed {α β : Type*} {R : α → β → Prop} | |
| {xs : List α} {ys : List β} (hr : List.Forall₂ R xs ys) : | |
| ∃ f : Fin xs.length → β, ys=List.ofFn f ∧ ∀ i,R (xs.get i) (f i) := by | |
| have hlen := hr.length_eq | |
| let f : Fin xs.length → β := fun i=>ys.get ⟨i.val,by rw [←hlen]; exact i.isLt⟩ | |
| refine ⟨f,?_,?_⟩ | |
| · apply List.ext_getElem | |
| · simp [hlen] | |
| · intro i hi hj | |
| simp [f] | |
| · intro i | |
| have hh := hr.get (i:=i.val) i.isLt (by rw [←hlen]; exact i.isLt) | |
| simpa [f,List.get_eq_getElem] using hh | |
| /-- Every physical pair retains its original row number and selected ID. -/ | |
| theorem matrixRel_typed {Rel : ℕ → ℕ → LayoutMatrixProgram.Pair → Prop} | |
| {B D : ℕ} {js : List ℕ} {rows : List (List LayoutMatrixProgram.Pair)} | |
| (hr : LayoutMatrixProgram.MatrixRel Rel B D js rows) : | |
| ∃ pairs : Fin D → Fin js.length → SignedWord, | |
| rows=List.ofFn (fun j=>List.ofFn (pairs j)) ∧ | |
| ∀ j i,Rel j.val (js.get i) (pairs j i) ∧ LayoutMatrixProgram.pairLength (pairs j i) ≤ B := by | |
| classical | |
| have hr' : List.Forall₂ (fun j row=>List.Forall₂ | |
| (fun v p=>Rel j v p ∧ LayoutMatrixProgram.pairLength p ≤ B) js row) | |
| (List.ofFn (fun j : Fin D=>j.val)) rows := by | |
| have he : List.ofFn (fun j : Fin D=>j.val)=List.range D := by | |
| apply List.ext_getElem | |
| · simp | |
| · intro i hi hj | |
| simp | |
| rw [he] | |
| exact hr | |
| obtain ⟨rr,hrr,hh⟩ := forall₂_typed hr' | |
| have hl : (List.ofFn (fun j : Fin D=>j.val)).length=D := by simp | |
| let rr' : Fin D → List LayoutMatrixProgram.Pair := fun j=>rr (Fin.cast hl.symm j) | |
| have he : rows=List.ofFn rr' := by simpa [rr'] using hrr | |
| have hrow (j : Fin D) : List.Forall₂ | |
| (fun v p=>Rel j.val v p ∧ LayoutMatrixProgram.pairLength p ≤ B) js (rr' j) := by | |
| simpa [rr'] using hh (Fin.cast hl.symm j) | |
| have hx : ∀ j,∃ f : Fin js.length → LayoutMatrixProgram.Pair, | |
| rr' j=List.ofFn f ∧ ∀ i,Rel j.val (js.get i) (f i) ∧ LayoutMatrixProgram.pairLength (f i) ≤ B := | |
| fun j=>forall₂_typed (hrow j) | |
| choose pairs hp hv using hx | |
| refine ⟨pairs,?_,hv⟩ | |
| exact he.trans (congrArg List.ofFn (funext hp)) | |
| /-- The coefficient matrix uses exactly the scanner's selected-ID order. -/ | |
| def selectedMatrix (C : ℕ → ℕ → ℤ) (D : ℕ) (js : List ℕ) : Fin D → Fin js.length → ℤ := | |
| fun j i=>C j.val (js.get i) | |
| /-- Exact cardinality of the actual output buckets, derived from the original | |
| entry relation and the physical preprocessing relation. -/ | |
| theorem actualBuckets_length_le_rank | |
| {Rel : ℕ → ℕ → LayoutMatrixProgram.Pair → Prop} {B D : ℕ} | |
| {js : List ℕ} {rows : List (List LayoutMatrixProgram.Pair)} | |
| (W : List Bool) {ys : List (List (List Bool))} (n H : ℕ) (C : ℕ → ℕ → ℤ) | |
| (hr : LayoutMatrixProgram.MatrixRel Rel B D js rows) | |
| (hy : List.Forall₂ (List.Forall₂ (shiftedWord W)) rows ys) | |
| (hrel : ∀ j v p,j < D→v∈js→Rel j v p→ | |
| SubexponentialSAT.BinaryArithmetic.toNat p.2 ≤ SubexponentialSAT.BinaryArithmetic.toNat W ∧ | |
| (SubexponentialSAT.BinaryArithmetic.toNat p.1 : ℤ)-(SubexponentialSAT.BinaryArithmetic.toNat p.2 : ℤ)=C j v) | |
| (hbound : ∀ j v p,j < D→v∈js→Rel j v p→|C j v| ≤ (H:ℤ)) : | |
| (BucketTupleProgram.actualBuckets W ys js n js.length).length ≤ | |
| (2*js.length*H+1)^LayoutRank.rowRank (selectedMatrix C D js) := by | |
| obtain ⟨pairs,rfl,hpairs⟩ := matrixRel_typed hr | |
| have hn (j : Fin D) (i : Fin js.length) := | |
| hrel j.val (js.get i) (pairs j i) j.isLt (List.get_mem ..) (hpairs j i).1 | |
| have hb (j : Fin D) (i : Fin js.length) : |selectedMatrix C D js j i| ≤ (H:ℤ) := | |
| hbound j.val (js.get i) (pairs j i) j.isLt (List.get_mem ..) (hpairs j i).1 | |
| have ht := storedTable_length_le_rank W pairs (selectedMatrix C D js) js n H hy | |
| (fun j i=>(hn j i).1) (fun j i=>(hn j i).2) hb | |
| simpa only [BucketTupleProgram.actualBuckets,BucketTupleProgram.controllerBuckets, | |
| SubexponentialSAT.Subexp.PackingBucketTable.table,List.length_map,Controller.labelledRecords, | |
| Controller.keys] using ht | |
| end SubexponentialSAT.SqrtLog.BucketProgram.ProducedRank | |
| end OAI.Algorithms | |