ZO-DARTS++: An Efficient and Size-Variable Zeroth-Order Neural Architecture Search Algorithm

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xie, Lunchen, Lomurno, Eugenio, Gambella, Matteo, Ardagna, Danilo, Roveri, Manual, Matteucci, Matteo, Shi, Qingjiang
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916646962069504
author Xie, Lunchen
Lomurno, Eugenio
Gambella, Matteo
Ardagna, Danilo
Roveri, Manual
Matteucci, Matteo
Shi, Qingjiang
author_facet Xie, Lunchen
Lomurno, Eugenio
Gambella, Matteo
Ardagna, Danilo
Roveri, Manual
Matteucci, Matteo
Shi, Qingjiang
contents Differentiable Neural Architecture Search (NAS) provides a promising avenue for automating the complex design of deep learning (DL) models. However, current differentiable NAS methods often face constraints in efficiency, operation selection, and adaptability under varying resource limitations. We introduce ZO-DARTS++, a novel NAS method that effectively balances performance and resource constraints. By integrating a zeroth-order approximation for efficient gradient handling, employing a sparsemax function with temperature annealing for clearer and more interpretable architecture distributions, and adopting a size-variable search scheme for generating compact yet accurate architectures, ZO-DARTS++ establishes a new balance between model complexity and performance. In extensive tests on medical imaging datasets, ZO-DARTS++ improves the average accuracy by up to 1.8\% over standard DARTS-based methods and shortens search time by approximately 38.6\%. Additionally, its resource-constrained variants can reduce the number of parameters by more than 35\% while maintaining competitive accuracy levels. Thus, ZO-DARTS++ offers a versatile and efficient framework for generating high-quality, resource-aware DL models suitable for real-world medical applications.
format Preprint
id arxiv_https___arxiv_org_abs_2503_06092
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ZO-DARTS++: An Efficient and Size-Variable Zeroth-Order Neural Architecture Search Algorithm
Xie, Lunchen
Lomurno, Eugenio
Gambella, Matteo
Ardagna, Danilo
Roveri, Manual
Matteucci, Matteo
Shi, Qingjiang
Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
I.5.1; I.5.4; I.2.6; I.2.10
Differentiable Neural Architecture Search (NAS) provides a promising avenue for automating the complex design of deep learning (DL) models. However, current differentiable NAS methods often face constraints in efficiency, operation selection, and adaptability under varying resource limitations. We introduce ZO-DARTS++, a novel NAS method that effectively balances performance and resource constraints. By integrating a zeroth-order approximation for efficient gradient handling, employing a sparsemax function with temperature annealing for clearer and more interpretable architecture distributions, and adopting a size-variable search scheme for generating compact yet accurate architectures, ZO-DARTS++ establishes a new balance between model complexity and performance. In extensive tests on medical imaging datasets, ZO-DARTS++ improves the average accuracy by up to 1.8\% over standard DARTS-based methods and shortens search time by approximately 38.6\%. Additionally, its resource-constrained variants can reduce the number of parameters by more than 35\% while maintaining competitive accuracy levels. Thus, ZO-DARTS++ offers a versatile and efficient framework for generating high-quality, resource-aware DL models suitable for real-world medical applications.
title ZO-DARTS++: An Efficient and Size-Variable Zeroth-Order Neural Architecture Search Algorithm
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
I.5.1; I.5.4; I.2.6; I.2.10
url https://arxiv.org/abs/2503.06092