Fine-block zero annihilates a singleton in any square-CW coarse component
Provedmme_dwz_arbitrary_coarse_component_singleton_zero_of_fine_blockcomponent-supportcoppersmith-winogradfine-gradingmatrix-multiplication
In an arbitrary coarse component of the squared Coppersmith--Winograd tensor, a selected lifted coarse-pair singleton is annihilated after arbitrary postmaps whenever the corresponding fine block tensor vanishes.
Preamble
import Definitions.Def_mme_dwz_step1_mixed_selected_full_map_data open MME Module PiTensorProduct open MME.DWZStep1Support open MME.DWZSourceAligned universe u set_option autoImplicit false
Formal statement
theorem mme_dwz_arbitrary_coarse_component_singleton_zero_of_fine_block
{K : Type u} [Field K] (coarse : Fin 3 → Fin 5)
(selected : ∀ i,
DWZComponentRestriction.LiftedCoarsePair.{u} 6 (coarse i))
{W : Fin 3 → Type u}
[∀ i, AddCommGroup (W i)] [∀ i, Module K (W i)]
(post : ∀ i,
((cwSquareCanonicalGrading K 6).blockSubtensor coarse).V i →ₗ[K]
W i)
(hzero : (cwSquareFineSplitGrading K 6).blockTensor
(fun i ↦ fineSplitGrade
(selected i).leftGrade (selected i).rightGrade) = 0) :
PiTensorProduct.map
(fun i ↦ (post i).comp
(DWZComponentRestriction.basisLabelProjection
(arbitraryCoarseComponentModeBasis K coarse i) id {selected i}))
((cwSquareCanonicalGrading K 6).blockSubtensor coarse).t = 0 := by
sorrySource
Duan--Wu--Zhou, Faster Matrix Multiplication via Asymmetric Hashing, arXiv:2210.10173v5, Section 6; https://arxiv.org/abs/2210.10173