Saved in:
Bibliographic Details
Main Authors: Zampini, Stefano, Zerbinati, Umberto, Turkiyyah, George, Keyes, David
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2403.12188
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909140192854016
author Zampini, Stefano
Zerbinati, Umberto
Turkiyyah, George
Keyes, David
author_facet Zampini, Stefano
Zerbinati, Umberto
Turkiyyah, George
Keyes, David
contents In recent years, we have witnessed the emergence of scientific machine learning as a data-driven tool for the analysis, by means of deep-learning techniques, of data produced by computational science and engineering applications. At the core of these methods is the supervised training algorithm to learn the neural network realization, a highly non-convex optimization problem that is usually solved using stochastic gradient methods. However, distinct from deep-learning practice, scientific machine-learning training problems feature a much larger volume of smooth data and better characterizations of the empirical risk functions, which make them suited for conventional solvers for unconstrained optimization. We introduce a lightweight software framework built on top of the Portable and Extensible Toolkit for Scientific computation to bridge the gap between deep-learning software and conventional solvers for unconstrained minimization. We empirically demonstrate the superior efficacy of a trust region method based on the Gauss-Newton approximation of the Hessian in improving the generalization errors arising from regression tasks when learning surrogate models for a wide range of scientific machine-learning techniques and test cases. All the conventional second-order solvers tested, including L-BFGS and inexact Newton with line-search, compare favorably, either in terms of cost or accuracy, with the adaptive first-order methods used to validate the surrogate models.
format Preprint
id arxiv_https___arxiv_org_abs_2403_12188
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle PETScML: Second-order solvers for training regression problems in Scientific Machine Learning
Zampini, Stefano
Zerbinati, Umberto
Turkiyyah, George
Keyes, David
Machine Learning
Mathematical Software
Optimization and Control
65K10, 68T07, 65M70, 65Y05
I.2.5; D.2.m; G.4; G.1.6; J.2
In recent years, we have witnessed the emergence of scientific machine learning as a data-driven tool for the analysis, by means of deep-learning techniques, of data produced by computational science and engineering applications. At the core of these methods is the supervised training algorithm to learn the neural network realization, a highly non-convex optimization problem that is usually solved using stochastic gradient methods. However, distinct from deep-learning practice, scientific machine-learning training problems feature a much larger volume of smooth data and better characterizations of the empirical risk functions, which make them suited for conventional solvers for unconstrained optimization. We introduce a lightweight software framework built on top of the Portable and Extensible Toolkit for Scientific computation to bridge the gap between deep-learning software and conventional solvers for unconstrained minimization. We empirically demonstrate the superior efficacy of a trust region method based on the Gauss-Newton approximation of the Hessian in improving the generalization errors arising from regression tasks when learning surrogate models for a wide range of scientific machine-learning techniques and test cases. All the conventional second-order solvers tested, including L-BFGS and inexact Newton with line-search, compare favorably, either in terms of cost or accuracy, with the adaptive first-order methods used to validate the surrogate models.
title PETScML: Second-order solvers for training regression problems in Scientific Machine Learning
topic Machine Learning
Mathematical Software
Optimization and Control
65K10, 68T07, 65M70, 65Y05
I.2.5; D.2.m; G.4; G.1.6; J.2
url https://arxiv.org/abs/2403.12188