English

Limits on representing Boolean functions by linear combinations of simple functions: thresholds, ReLUs, and low-degree polynomials

Computational Complexity 2018-02-27 v1 Discrete Mathematics Neural and Evolutionary Computing

Abstract

We consider the problem of representing Boolean functions exactly by "sparse" linear combinations (over R\mathbb{R}) of functions from some "simple" class C{\cal C}. In particular, given C{\cal C} we are interested in finding low-complexity functions lacking sparse representations. When C{\cal C} is the set of PARITY functions or the set of conjunctions, this sort of problem has a well-understood answer, the problem becomes interesting when C{\cal C} is "overcomplete" and the set of functions is not linearly independent. We focus on the cases where C{\cal C} is the set of linear threshold functions, the set of rectified linear units (ReLUs), and the set of low-degree polynomials over a finite field, all of which are well-studied in different contexts. We provide generic tools for proving lower bounds on representations of this kind. Applying these, we give several new lower bounds for "semi-explicit" Boolean functions. For example, we show there are functions in nondeterministic quasi-polynomial time that require super-polynomial size: \bullet Depth-two neural networks with sign activation function, a special case of depth-two threshold circuit lower bounds. \bullet Depth-two neural networks with ReLU activation function. \bullet R\mathbb{R}-linear combinations of O(1)O(1)-degree Fp\mathbb{F}_p-polynomials, for every prime pp (related to problems regarding Higher-Order "Uncertainty Principles"). We also obtain a function in ENPE^{NP} requiring 2Ω(n)2^{\Omega(n)} linear combinations. \bullet R\mathbb{R}-linear combinations of ACCTHRACC \circ THR circuits of polynomial size (further generalizing the recent lower bounds of Murray and the author). (The above is a shortened abstract. For the full abstract, see the paper.)

Keywords

Cite

@article{arxiv.1802.09121,
  title  = {Limits on representing Boolean functions by linear combinations of simple functions: thresholds, ReLUs, and low-degree polynomials},
  author = {R. Ryan Williams},
  journal= {arXiv preprint arXiv:1802.09121},
  year   = {2018}
}
R2 v1 2026-06-23T00:33:00.119Z