English

Approximation with one-bit polynomials in Bernstein form

Information Theory 2022-12-08 v2 Classical Analysis and ODEs math.IT

Abstract

We prove various theorems on approximation using polynomials with integer coefficients in the Bernstein basis of any given order. In the extreme, we draw the coefficients from {±1}\{ \pm 1\} only. A basic case of our results states that for any Lipschitz function f:[0,1][1,1]f:[0,1] \to [-1,1] and for any positive integer nn, there are signs σ0,,σn{±1}\sigma_0,\dots,\sigma_n \in \{\pm 1\} such that f(x)k=0nσk(nk)xk(1x)nkC(1+fLip)1+nx(1x) \mboxforallx[0,1].\left |f(x) - \sum_{k=0}^n \sigma_k \, \binom{n}{k} x^k (1-x)^{n-k} \right | \leq \frac{C (1+|f|_{\mathrm{Lip}})}{1+\sqrt{nx(1-x)}} ~\mbox{ for all } x \in [0,1]. More generally, we show that higher accuracy is achievable for smoother functions: For any integer s1s\geq 1, if ff has a Lipschitz (s1)(s{-}1)st derivative, then approximation accuracy of order O(ns/2)O(n^{-s/2}) is achievable with coefficients in {±1}\{\pm 1\} provided f<1\|f \|_\infty < 1, and of order O(ns)O(n^{-s}) with unrestricted integer coefficients, both uniformly on closed subintervals of (0,1)(0,1) as above. Hence these polynomial approximations are not constrained by the saturation of classical Bernstein polynomials. Our approximations are constructive and can be implemented using feedforward neural networks whose weights are chosen from {±1}\{\pm 1\} only.

Keywords

Cite

@article{arxiv.2112.09183,
  title  = {Approximation with one-bit polynomials in Bernstein form},
  author = {C. Sinan Güntürk and Weilin Li},
  journal= {arXiv preprint arXiv:2112.09183},
  year   = {2022}
}

Comments

28 pages

R2 v1 2026-06-24T08:21:07.638Z