Tensor extraction onto the whole six-region interface
Provedmme_released_global_joint_window_extractionmatrix-multiplicationmore-asymmetrytensor-restriction
Let be six physical global Parts at one scale, with input counts and logarithmic output rates . Let be the published joint-window predicate obtained by imposing each regional window on its own positions. Over every field, there exists an integer such that
Here is the common regional block count and denotes actual tensor restriction. This is the whole interface on which a joint recursive recipe can act.
Preamble
import Definitions.Def_mme_released_global_joint_interface open BigOperators MME MME.TensorObj MME.ProfiledCW MME.GlobalCW MME.RegionRealization MME.ReleasedGlobal set_option autoImplicit false universe u
Formal statement
theorem mme_released_global_joint_window_extraction {K : Type u} [Field K] (k : ℕ) (hk : 0 < k)
(a : ∀ o : Fin 6, Reference o k) (eps : Fin 6 → ℝ)
(S : ∀ o, Part (4 * blocks k) 3 (physicalWindow o k hk (a o) (eps o))) :
∃ outputs : ℕ, Real.exp (∑ o, (S o).rate) ≤ (outputs : ℝ) ∧
Restrict (bigAdd (fun _ : Fin outputs ↦ tensor K (jointWindow k hk a eps)))
(bigAdd (fun _ : Fin (∏ o, (S o).inputs) ↦
(CWObj K 5).kronPow (4 * (6 * blocks k)))) := by sorry
Source
Auxiliary formalization for Alman et al., More Asymmetry Yields Faster Matrix Multiplication, https://arxiv.org/html/2404.16349v2, Theorem 5.3, Section 5.1, Theorem 6.4 and Algorithm 1. Specialization to the published exact ReleasedGlobal seed; the numerical recursive continuation remains an explicit separate obligation.