English

Faster Online Matrix-Vector Multiplication

Data Structures and Algorithms 2016-05-09 v2 Computational Complexity

Abstract

We consider the Online Boolean Matrix-Vector Multiplication (OMV) problem studied by Henzinger et al. [STOC'15]: given an n×nn \times n Boolean matrix MM, we receive nn Boolean vectors v1,,vnv_1,\ldots,v_n one at a time, and are required to output MviM v_i (over the Boolean semiring) before seeing the vector vi+1v_{i+1}, for all ii. Previous known algorithms for this problem are combinatorial, running in O(n3/log2n)O(n^3/\log^2 n) time. Henzinger et al. conjecture there is no O(n3ε)O(n^{3-\varepsilon}) time algorithm for OMV, for all ε>0\varepsilon > 0; their OMV conjecture is shown to imply strong hardness results for many basic dynamic problems. We give a substantially faster method for computing OMV, running in n3/2Ω(logn)n^3/2^{\Omega(\sqrt{\log n})} randomized time. In fact, after seeing 2ω(logn)2^{\omega(\sqrt{\log n})} vectors, we already achieve n2/2Ω(logn)n^2/2^{\Omega(\sqrt{\log n})} amortized time for matrix-vector multiplication. Our approach gives a way to reduce matrix-vector multiplication to solving a version of the Orthogonal Vectors problem, which in turn reduces to "small" algebraic matrix-matrix multiplication. Applications include faster independent set detection, partial match retrieval, and 2-CNF evaluation. We also show how a modification of our method gives a cell probe data structure for OMV with worst case O(n7/4/w)O(n^{7/4}/\sqrt{w}) time per query vector, where ww is the word size. This result rules out an unconditional proof of the OMV conjecture using purely information-theoretic arguments.

Keywords

Cite

@article{arxiv.1605.01695,
  title  = {Faster Online Matrix-Vector Multiplication},
  author = {Kasper Green Larsen and Ryan Williams},
  journal= {arXiv preprint arXiv:1605.01695},
  year   = {2016}
}
R2 v1 2026-06-22T13:54:09.894Z