FUSE: First-Order and Second-Order Unified SynthEsis in Stochastic Optimization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Jiang, Zhanhong, Hasan, Md Zahid, Balu, Aditya, Waite, Joshua R., Huang, Genyi, Sarkar, Soumik
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916645169004544
author Jiang, Zhanhong
Hasan, Md Zahid
Balu, Aditya
Waite, Joshua R.
Huang, Genyi
Sarkar, Soumik
author_facet Jiang, Zhanhong
Hasan, Md Zahid
Balu, Aditya
Waite, Joshua R.
Huang, Genyi
Sarkar, Soumik
contents Stochastic optimization methods have actively been playing a critical role in modern machine learning algorithms to deliver decent performance. While numerous works have proposed and developed diverse approaches, first-order and second-order methods are in entirely different situations. The former is significantly pivotal and dominating in emerging deep learning but only leads convergence to a stationary point. However, second-order methods are less popular due to their computational intensity in large-dimensional problems. This paper presents a novel method that leverages both the first-order and second-order methods in a unified algorithmic framework, termed FUSE, from which a practical version (PV) is derived accordingly. FUSE-PV stands as a simple yet efficient optimization method involving a switch-over between first and second orders. Additionally, we develop different criteria that determine when to switch. FUSE-PV has provably shown a smaller computational complexity than SGD and Adam. To validate our proposed scheme, we present an ablation study on several simple test functions and show a comparison with baselines for benchmark datasets.
format Preprint
id arxiv_https___arxiv_org_abs_2503_04204
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle FUSE: First-Order and Second-Order Unified SynthEsis in Stochastic Optimization
Jiang, Zhanhong
Hasan, Md Zahid
Balu, Aditya
Waite, Joshua R.
Huang, Genyi
Sarkar, Soumik
Computer Vision and Pattern Recognition
Machine Learning
Stochastic optimization methods have actively been playing a critical role in modern machine learning algorithms to deliver decent performance. While numerous works have proposed and developed diverse approaches, first-order and second-order methods are in entirely different situations. The former is significantly pivotal and dominating in emerging deep learning but only leads convergence to a stationary point. However, second-order methods are less popular due to their computational intensity in large-dimensional problems. This paper presents a novel method that leverages both the first-order and second-order methods in a unified algorithmic framework, termed FUSE, from which a practical version (PV) is derived accordingly. FUSE-PV stands as a simple yet efficient optimization method involving a switch-over between first and second orders. Additionally, we develop different criteria that determine when to switch. FUSE-PV has provably shown a smaller computational complexity than SGD and Adam. To validate our proposed scheme, we present an ablation study on several simple test functions and show a comparison with baselines for benchmark datasets.
title FUSE: First-Order and Second-Order Unified SynthEsis in Stochastic Optimization
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2503.04204