DocumentCode :
3018912
Title :
Approximating Boolean Functions with Depth-2 Circuits
Author :
Blais, Eric ; Li-Yang Tan
Author_Institution :
CSAIL, MIT, Cambridge, MA, USA
fYear :
2013
fDate :
5-7 June 2013
Firstpage :
74
Lastpage :
85
Abstract :
We study the complexity of approximating Boolean functions with DNFs and other depth-2 circuits, exploring two main directions: universal bounds on the approximability of all Boolean functions, and the approximability of the parity function. In the first direction, our main positive results are the first non-trivial universal upper bounds on appropriability by DNFs: : Every Boolean function can be ε-approximated by a DNF of size Oε(2n/ log n). : Every Boolean function can be ε-approximated by a DNF of width cε n, where cε <; 1. Our techniques extend broadly to give strong universal upper bounds on approximability by various depth-2 circuits that generalize DNFs, including the intersection of halfspaces, low-degree PTFs, and unate functions. We show that the parameters of our constructions come close to matching the information-theoretic inapproximability of a random function. In the second direction our main positive result is the construction of an explicit DNF that approximates the parity function: : PARn can be ε-approximated by a DNF of size 2(1-2ε)n and width (1 - 2ε)n. Using Fourier analytic tools we show that our construction is essentially optimal not just within the class of DNFs, but also within the far more expressive classes of the intersection of halfspaces and intersection of unate functions.
Keywords :
Boolean functions; Fourier analysis; approximation theory; random functions; Boolean functions approximation; Fourier analytic tools; depth-2 circuits; explicit DNF; information-theoretic inapproximability; low-degree PTF; nontrivial universal upper bounds; parity function; random function; unate functions; Approximation methods; Boolean functions; Complexity theory; Computational modeling; Hypercubes; Noise; Upper bound; Kuznetsov Theorem ([Kor83]; The optimal DNF size for a random Boolean function is (K + o(1))(2n= log n log log n); [Kuz83]1); where <K<1:54169.;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Computational Complexity (CCC), 2013 IEEE Conference on
Conference_Location :
Stanford, CA
Type :
conf
DOI :
10.1109/CCC.2013.17
Filename :
6597751
Link To Document :
بازگشت