Smooth Approximations of the Rounding Function

Fuente: arXiv
Saved in:
Bibliographic Details
Main Author: Semenov, Stanislav
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918001404542976
author Semenov, Stanislav
author_facet Semenov, Stanislav
contents We propose novel smooth approximations to the classical rounding function, suitable for differentiable optimization and machine learning applications. Our constructions are based on two approaches: (1) localized sigmoid window functions centered at each integer, and (2) normalized weighted sums of sigmoid derivatives representing local densities. The first method approximates the step-like behavior of rounding through differences of shifted sigmoids, while the second method achieves smooth interpolation between integers via density-based weighting. Both methods converge pointwise to the classical rounding function as the sharpness parameter k tends to infinity, and allow controlled trade-offs between smoothness and approximation accuracy. We demonstrate that by restricting the summation to a small set of nearest integers, the computational cost remains low without sacrificing precision. These constructions provide fully differentiable alternatives to hard rounding, which are valuable in contexts where gradient-based methods are essential.
format Preprint
id arxiv_https___arxiv_org_abs_2504_19026
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Smooth Approximations of the Rounding Function
Semenov, Stanislav
Machine Learning
Optimization and Control
03F60, 26E40
F.4.1; F.1.1
We propose novel smooth approximations to the classical rounding function, suitable for differentiable optimization and machine learning applications. Our constructions are based on two approaches: (1) localized sigmoid window functions centered at each integer, and (2) normalized weighted sums of sigmoid derivatives representing local densities. The first method approximates the step-like behavior of rounding through differences of shifted sigmoids, while the second method achieves smooth interpolation between integers via density-based weighting. Both methods converge pointwise to the classical rounding function as the sharpness parameter k tends to infinity, and allow controlled trade-offs between smoothness and approximation accuracy. We demonstrate that by restricting the summation to a small set of nearest integers, the computational cost remains low without sacrificing precision. These constructions provide fully differentiable alternatives to hard rounding, which are valuable in contexts where gradient-based methods are essential.
title Smooth Approximations of the Rounding Function
topic Machine Learning
Optimization and Control
03F60, 26E40
F.4.1; F.1.1
url https://arxiv.org/abs/2504.19026