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

10/16/2017
by   Suryajith Chillara, et al.
0

In this paper, we study the algebraic formula complexity of multiplying d many 2× 2 matrices, denoted 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 Δ≤ d, we show that any product-depth Δ multilinear formula for IMM_d must have size (Ω(Δ d^1/Δ)). It also follows from this that any multilinear circuit of product-depth Δ for the same polynomial of the above form must have a size of (Ω(d^1/Δ)). In particular, any polynomial-sized multilinear formula for IMM_d must have depth Ω( d), and any polynomial-sized multilinear circuit for IMM_d must have depth Ω( d/ 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 s can be converted to one of size s^O(1) and depth O( 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( s) without a superpolynomial blow-up in size. 2. Separations from general formulas: Our result, along with a non-trivial upper bound for IMM_d implied by a result of Gupta, Kamath, Kayal and Saptharishi (SICOMP 2016), shows that for any size s and product-depth Δ = o( s), general formulas of size s and product-depth Δ cannot be converted to multilinear formulas of size s^ω(1) and product-depth Δ, when the underlying field has characteristic zero.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
05/02/2022

Improved Low-Depth Set-Multilinear Circuit Lower Bounds

We prove strengthened lower bounds for constant-depth set-multilinear fo...
research
02/14/2023

Towards Optimal Depth-Reductions for Algebraic Formulas

Classical results of Brent, Kuck and Maruyama (IEEE Trans. Computers 197...
research
04/28/2020

Tree-depth and the Formula Complexity of Subgraph Isomorphism

For a fixed "pattern" graph G, the colored G-subgraph isomorphism proble...
research
06/13/2018

Subspace-Invariant AC^0 Formulas

We consider the action of a linear subspace U of {0,1}^n on the set of A...
research
02/19/2019

Towards Optimal Depth Reductions for Syntactically Multilinear Circuits

We show that any n-variate polynomial computable by a syntactically mult...
research
06/21/2023

Trotter error bounds and dynamic multi-product formulas for Hamiltonian simulation

Multi-product formulas are linear combinations of Trotter circuits offer...
research
10/09/2017

Discovering the roots: Uniform closure results for algebraic classes under factoring

Newton iteration (NI) is an almost 350 years old recursive formula that ...

Please sign up or login with your details

Forgot password? Click here to reset