Volume 2 (2006) Article 6 pp. 121-135

Separation of Multilinear Circuit and Formula Size

by Ran Raz

Received: September 30, 2005
Revised: March 4, 2006
Published: May 2, 2006
DOI: 10.4086/toc.2006.v002a006

Download article from ToC site:
[PS (288K)]    [PS.GZ (88K)]    [PS.ZIP (88K)]
[PDF (161K)]    [DVI (120K)]    [TeX (45K)]
Misc.:
[HTML Bibliography] [Bibliography Source] [BibTeX] 
[About the Author(s)]
Keywords: arithmetic circuits, log-depth, multilinear polynomials, lower bounds, separation of complexity classes
Categories: complexity theory, lower bounds, circuits, arithmetic circuits, formulas, arithmetic formulas, complexity classes, separation of complexity classes, polynomials
ACM Classification: F.2.2, F.1.3, F.1.2, G.2.0
AMS Classification: 68Q15, 68Q17, 68Q25, 94C05

Abstract: [Plain Text Version]

An arithmetic circuit or formula is multilinear if the polynomial computed at each of its wires is multilinear. We give an explicit polynomial f(x1,...,xn) with coefficients in {0,1} such that over any field: This gives a super-polynomial gap between multilinear circuit and formula size, and separates multilinear NC1 circuits from multilinear NC2 circuits.