English

Small-depth Multilinear Formula Lower Bounds for Iterated Matrix Multiplication, with Applications

Computational Complexity 2017-10-17 v1

Abstract

In this paper, we study the algebraic formula complexity of multiplying dd many 2×22\times 2 matrices, denoted IMMd\mathrm{IMM}_{d}, and show that the well-known divide-and-conquer algorithm cannot be significantly improved at any depth, as long as the formulas are multilinear. Formally, for each depth Δlogd\Delta \leq \log d, we show that any product-depth Δ\Delta multilinear formula for IMMd\mathrm{IMM}_d must have size exp(Ω(Δd1/Δ)).\exp(\Omega(\Delta d^{1/\Delta})). It also follows from this that any multilinear circuit of product-depth Δ\Delta for the same polynomial of the above form must have a size of exp(Ω(d1/Δ)).\exp(\Omega(d^{1/\Delta})). In particular, any polynomial-sized multilinear formula for IMMd\mathrm{IMM}_d must have depth Ω(logd)\Omega(\log d), and any polynomial-sized multilinear circuit for IMMd\mathrm{IMM}_d must have depth Ω(logd/loglogd).\Omega(\log d/\log \log d). Both these bounds are tight up to constant factors. 1. Depth-reduction: A well-known result of Brent (JACM 1974) implies that any formula of size ss can be converted to one of size sO(1)s^{O(1)} and depth O(logs)O(\log s); further, this reduction continues to hold for multilinear formulas. Our lower bound implies that any depth-reduction in the multilinear setting cannot reduce the depth to o(logs)o(\log s) without a superpolynomial blow-up in size. 2. Separations from general formulas: Our result, along with a non-trivial upper bound for IMMd\mathrm{IMM}_{d} implied by a result of Gupta, Kamath, Kayal and Saptharishi (SICOMP 2016), shows that for any size ss and product-depth Δ=o(logs),\Delta = o(\log s), general formulas of size ss and product-depth Δ\Delta cannot be converted to multilinear formulas of size sω(1)s^{\omega(1)} and product-depth Δ,\Delta, when the underlying field has characteristic zero.

Keywords

Cite

@article{arxiv.1710.05481,
  title  = {Small-depth Multilinear Formula Lower Bounds for Iterated Matrix Multiplication, with Applications},
  author = {Suryajith Chillara and Nutan Limaye and Srikanth Srinivasan},
  journal= {arXiv preprint arXiv:1710.05481},
  year   = {2017}
}
R2 v1 2026-06-22T22:14:24.599Z