Weizmann Logo
ECCC
Electronic Colloquium on Computational Complexity

Under the auspices of the Computational Complexity Foundation (CCF)

Login | Register | Classic Style



REPORTS > DETAIL:

Revision(s):

Revision #1 to TR19-067 | 2nd May 2022 18:14

Sign rank vs Discrepancy

RSS-Feed




Revision #1
Authors: Hamed Hatami, Kaave Hosseini, Shachar Lovett
Accepted on: 2nd May 2022 18:14
Downloads: 343
Keywords: 


Abstract:

Sign-rank and discrepancy are two central notions in communication complexity. The seminal work of Babai, Frankl, and Simon from 1986 initiated an active line of research that investigates the gap between these two notions.
In this article, we establish the strongest possible separation by constructing a boolean matrix whose sign-rank is only $3$, and yet its discrepancy is $2^{-\Omega(n)}$. We note that every matrix of sign-rank $2$ has discrepancy $n^{-O(1)}$.

In connection to learning theory, our result implies the existence of Boolean matrices whose entries are represented by points and half-spaces in dimension $3$, and yet, the normalized margin of any such representation, even in higher dimensions, is bound to be very small.

In the context of communication complexity, our result in particular implies that there are boolean functions with $O(1)$ unbounded-error randomized communication complexity while having $\Omega(n)$ weakly unbounded-error randomized communication complexity.


Paper:

TR19-067 | 6th May 2019 19:23

Sign rank vs Discrepancy


Abstract:

Sign-rank and discrepancy are two central notions in communication complexity. The seminal work of Babai, Frankl, and Simon from 1986 initiated an active line of research that investigates the gap between these two notions.
In this article, we establish the strongest possible separation by constructing a Boolean matrix whose sign-rank is only $3$, and yet its discrepancy is $2^{-\Omega(n)}$. We note that every matrix of sign-rank $2$ has discrepancy $n^{-O(1)}$.
Our result in particular implies that there are Boolean functions with $O(1)$ unbounded error randomized communication complexity while having $\Omega(n)$ weakly unbounded error randomized communication complexity.



ISSN 1433-8092 | Imprint