Stochastic Penalty-Barrier Methods for Constrained Machine Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Bosák, Adam, Kliachkin, Andrii, Lepšová, Jana, Bareilles, Gilles, Mareček, Jakub
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913146092912640
author Bosák, Adam
Kliachkin, Andrii
Lepšová, Jana
Bareilles, Gilles
Mareček, Jakub
author_facet Bosák, Adam
Kliachkin, Andrii
Lepšová, Jana
Bareilles, Gilles
Mareček, Jakub
contents Constrained machine learning enables fairness-aware training, physics-informed neural networks, and integration of symbolic domain knowledge into statistical models. Despite its practical importance, no general method exists for the non-convex, non-smooth, stochastic setting that arises naturally in deep learning. We propose the Stochastic Penalty-Barrier Method (SPBM), which extends classical penalty and barrier methods to this setting via exponential dual averaging, a stabilized penalty schedule, and the Moreau envelope to handle non-smoothness. Experiments across multiple settings show that SPBM matches or outperforms existing constrained optimization baselines while incurring only linear runtime overhead compared to unconstrained Adam for up to 10,000 constraints.
format Preprint
id arxiv_https___arxiv_org_abs_2605_18618
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Stochastic Penalty-Barrier Methods for Constrained Machine Learning
Bosák, Adam
Kliachkin, Andrii
Lepšová, Jana
Bareilles, Gilles
Mareček, Jakub
Machine Learning
Artificial Intelligence
Constrained machine learning enables fairness-aware training, physics-informed neural networks, and integration of symbolic domain knowledge into statistical models. Despite its practical importance, no general method exists for the non-convex, non-smooth, stochastic setting that arises naturally in deep learning. We propose the Stochastic Penalty-Barrier Method (SPBM), which extends classical penalty and barrier methods to this setting via exponential dual averaging, a stabilized penalty schedule, and the Moreau envelope to handle non-smoothness. Experiments across multiple settings show that SPBM matches or outperforms existing constrained optimization baselines while incurring only linear runtime overhead compared to unconstrained Adam for up to 10,000 constraints.
title Stochastic Penalty-Barrier Methods for Constrained Machine Learning
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2605.18618