Saved in:
Bibliographic Details
Main Authors: Choi, Soobin, Cepeda, Valentina, Gomez, Andres, Han, Shaoning
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2504.16330
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916702902550528
author Choi, Soobin
Cepeda, Valentina
Gomez, Andres
Han, Shaoning
author_facet Choi, Soobin
Cepeda, Valentina
Gomez, Andres
Han, Shaoning
contents We investigate convexification for convex quadratic optimization with step function penalties. Such problems can be cast as mixed-integer quadratic optimization problems, where binary variables are used to encode the non-convex step function. First, we derive the convex hull for the epigraph of a quadratic function defined by a rank-one matrix and step function penalties. Using this rank-one convexification, we develop copositive and semi-definite relaxations for general convex quadratic functions. Leveraging these findings, we construct convex formulations to the support vector machine problem with 0--1 loss and show that they yield robust estimators in settings with anomalies and outliers.
format Preprint
id arxiv_https___arxiv_org_abs_2504_16330
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Rank-one convexification for quadratic optimization problems with step function penalties
Choi, Soobin
Cepeda, Valentina
Gomez, Andres
Han, Shaoning
Optimization and Control
90
We investigate convexification for convex quadratic optimization with step function penalties. Such problems can be cast as mixed-integer quadratic optimization problems, where binary variables are used to encode the non-convex step function. First, we derive the convex hull for the epigraph of a quadratic function defined by a rank-one matrix and step function penalties. Using this rank-one convexification, we develop copositive and semi-definite relaxations for general convex quadratic functions. Leveraging these findings, we construct convex formulations to the support vector machine problem with 0--1 loss and show that they yield robust estimators in settings with anomalies and outliers.
title Rank-one convexification for quadratic optimization problems with step function penalties
topic Optimization and Control
90
url https://arxiv.org/abs/2504.16330