TY - JOUR
T1 - On the Structure of Boolean Functions with Small Spectral Norm
AU - Shpilka, Amir
AU - Tal, Avishay
AU - Volk, Ben lee
N1 - Publisher Copyright:
© 2015, Springer Basel.
PY - 2017/3/1
Y1 - 2017/3/1
N2 - In this paper we prove results regarding Boolean functions with small spectral norm (the spectral norm of f is ‖ f^ ‖ 1= ∑ α| f^ (α) |). Specifically, we prove the following results for functions f: { 0 , 1 } n→ { 0 , 1 } with ‖ f^ ‖ 1= A.1.There is an affine subspace V of co-dimension at most A2 such that f| V is constant.2.f can be computed by a parity decision tree of size at most 2A2n2A. (A parity decision tree is a decision tree whose nodes are labeled with arbitrary linear functions.)3.f can be computed by a De Morgan formula of size O(2A2n2A+2) and by a De Morgan formula of depth O(A2+ log (n) · A).4.If in addition f has at most s nonzero Fourier coefficients, then f can be computed by a parity decision tree of depth at most A2log s.5.For every ϵ> 0 there is a parity decision tree of depth O(A2+ log (1 / ϵ)) and size 2O(A2)·min{1/ϵ2,log(1/ϵ)2A} that ϵ-approximates f. Furthermore, this tree can be learned (in the uniform distribution model), with probability 1 - δ, using poly(n, exp (A2) , 1 / ϵ, log (1 / δ)) membership queries.All the results above (except 3) also hold (with a slight change in parameters) for functions f:Zpn→{0,1}.
AB - In this paper we prove results regarding Boolean functions with small spectral norm (the spectral norm of f is ‖ f^ ‖ 1= ∑ α| f^ (α) |). Specifically, we prove the following results for functions f: { 0 , 1 } n→ { 0 , 1 } with ‖ f^ ‖ 1= A.1.There is an affine subspace V of co-dimension at most A2 such that f| V is constant.2.f can be computed by a parity decision tree of size at most 2A2n2A. (A parity decision tree is a decision tree whose nodes are labeled with arbitrary linear functions.)3.f can be computed by a De Morgan formula of size O(2A2n2A+2) and by a De Morgan formula of depth O(A2+ log (n) · A).4.If in addition f has at most s nonzero Fourier coefficients, then f can be computed by a parity decision tree of depth at most A2log s.5.For every ϵ> 0 there is a parity decision tree of depth O(A2+ log (1 / ϵ)) and size 2O(A2)·min{1/ϵ2,log(1/ϵ)2A} that ϵ-approximates f. Furthermore, this tree can be learned (in the uniform distribution model), with probability 1 - δ, using poly(n, exp (A2) , 1 / ϵ, log (1 / δ)) membership queries.All the results above (except 3) also hold (with a slight change in parameters) for functions f:Zpn→{0,1}.
KW - Analysis of Boolean functions
KW - parity decision trees
KW - spectral norm
UR - http://www.scopus.com/inward/record.url?scp=84940507764&partnerID=8YFLogxK
U2 - 10.1007/s00037-015-0110-y
DO - 10.1007/s00037-015-0110-y
M3 - ???researchoutput.researchoutputtypes.contributiontojournal.article???
AN - SCOPUS:84940507764
SN - 1016-3328
VL - 26
SP - 229
EP - 273
JO - Computational Complexity
JF - Computational Complexity
IS - 1
ER -