New results in canonical polyadic decomposition overfinite fields

Abstract

Canonical polyadic decomposition (CPD) consists of expressing a tensor (multidimensional array) as a sum of several rank-1 tensors, each of which is an outer/separable product of vectors. The number of rank-1 tensors used in a CPD is called the rank of the CPD, and the minimum possible rank of a CPD for a given tensor is called the rank of the tensor. CPD is at the core of fast matrix multiplication, a computational problem with widespread implications across several seemingly unrelated problems in computer science. Much recent progress in this field has used randomized heuristic search to find new CPDs, often over a finite field. However, if these techniques fail to find a CPD with low enough rank, they cannot prove that no such CPD exists. Consequently, these methods fail to resolve certain long-standing questions, such as whether the tensor corresponding to 3 × 3 matrix multiplication has rank less than 23. To make progress on these problems, we develop a novel algorithm that preserves exactness, i.e. they can provably verify whether or not a given tensor has a specified rank. Compared to brute force, when searching for a rank-R CPD of a n0 × · · · × nD−1-shaped tensor over a finite field F, where n0 ≥ · · · ≥ nD−1, our algorithm saves a multiplicative factor of roughly |F| R(n0−1)+n0( P d≥1 nd) . Additionally, our algorithm runs in polynomial time. We also find a novel algorithm to search border CPDs, a variant of CPDs that is also important in fast matrix multiplication. Finally, we study the maximum rank problem and give new upper and lower bounds, both for families of tensor shapes and specific shapes. Although our CPD search algorithms are still too slow to resolve the rank of 3 × 3 matrix multiplication, we are able to utilize them in this problem by adding extra search pruners that do not affect exactness or increase asymptotic running time.