Uniqueness of Optimal Mod 3 Circuits for Parity

Authors Frederic Green, Amitabha Roy



PDF
Thumbnail PDF

File

DagSemProc.07411.7.pdf
  • Filesize: 213 kB
  • 15 pages

Document Identifiers

Author Details

Frederic Green
Amitabha Roy

Cite AsGet BibTex

Frederic Green and Amitabha Roy. Uniqueness of Optimal Mod 3 Circuits for Parity. In Algebraic Methods in Computational Complexity. Dagstuhl Seminar Proceedings, Volume 7411, pp. 1-15, Schloss Dagstuhl – Leibniz-Zentrum für Informatik (2008)
https://doi.org/10.4230/DagSemProc.07411.7

Abstract

We prove that the quadratic polynomials modulo $3$ with the largest correlation with parity are unique up to permutation of variables and constant factors. As a consequence of our result, we completely characterize the smallest MAJ~$circ mbox{MOD}_3 circ { m AND}_2$ circuits that compute parity, where a MAJ~$circ mbox{MOD}_3 circ { m AND}_2$ circuit is one that has a majority gate as output, a middle layer of MOD$_3$ gates and a bottom layer of AND gates of fan-in $2$. We also prove that the sub-optimal circuits exhibit a stepped behavior: any sub-optimal circuits of this class that compute parity must have size at least a factor of $frac{2}{sqrt{3}}$ times the optimal size. This verifies, for the special case of $m=3$, two conjectures made by Due~{n}ez, Miller, Roy and Straubing (Journal of Number Theory, 2006) for general MAJ~$circ mathrm{MOD}_m circ { m AND}_2$ circuits for any odd $m$. The correlation and circuit bounds are obtained by studying the associated exponential sums, based on some of the techniques developed by Green (JCSS, 2004). We regard this as a step towards obtaining tighter bounds both for the $m ot = 3$ quadratic case as well as for higher degrees.
Keywords
  • Circuit complexity
  • correlations
  • exponential sums

Metrics

  • Access Statistics
  • Total Accesses (updated on a weekly basis)
    0
    PDF Downloads
Questions / Remarks / Feedback
X

Feedback for Dagstuhl Publishing


Thanks for your feedback!

Feedback submitted

Could not send message

Please try again later or send an E-mail