Self-scaled bounds for atomic cone ranks: applications to nonnegative rank and cp-rank
Name
10107_2015_937_ReferencePDF.pdf
Size
525.23 KB
Format
Adobe PDF
Checksum (MD5)
27cb0ff04f3428ce1e7c3c8633502678
Author(s)
Fawzi, Hamza
Parrilo, Pablo A.
Date Issued
August 2015
Journal
Mathematical Programming
Publisher
Springer Berlin Heidelberg
Citation
Fawzi, Hamza, and Pablo A. Parrilo. “Self-Scaled Bounds for Atomic Cone Ranks: Applications to Nonnegative Rank and Cp-Rank.” Mathematical Programming 158.1–2 (2016): 417–465.
Version
Author's final manuscript
Abstract
The nonnegative rank of a matrix A is the smallest integer r such that A can be written as the sum of r rank-one nonnegative matrices. The nonnegative rank has received a lot of attention recently due to its application in optimization, probability and communication complexity. In this paper we study a class of atomic rank functions defined on a convex cone which generalize several notions of “positive” ranks such as nonnegative rank or cp-rank (for completely positive matrices). The main contribution of the paper is a new method to obtain lower bounds for such ranks. Additionally the bounds we propose can be computed by semidefinite programming using sum-of-squares relaxations. The idea of the lower bound relies on an atomic norm approach where the atoms are self-scaled according to the vector (or matrix, in the case of nonnegative rank) of interest. This results in a lower bound that is invariant under scaling and that enjoys other interesting structural properties. For the case of the nonnegative rank we show that our bound has an appealing connection with existing combinatorial bounds and other norm-based bounds. For example we show that our lower bound is a non-combinatorial version of the fractional rectangle cover number, while the sum-of-squares relaxation is closely related to the Lovász [bar over ϑ] number of the rectangle graph of the matrix. We also prove that the lower bound is always greater than or equal to the hyperplane separation bound (and other similar “norm-based” bounds). We also discuss the case of the tensor nonnegative rank as well as the cp-rank, and compare our bound with existing results.
MIT Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Massachusetts Institute of Technology. Laboratory for Information and Decision Systems
Terms of Use
Creative Commons Attribution-Noncommercial-Share Alike
Persistent DSpace Link
DOI of Published Version
https://doi.org/10.1007/s10107-015-0937-7
