F. A. Buzaev, D. S. Efremenko, I. A. Chuprov, Ya. N. Khassan, E. N. Kazakov, J. Gao
{"title":"Search for Optimal Architecture of Physics-Informed Neural Networks Using Differential Evolution Algorithm","authors":"F. A. Buzaev, D. S. Efremenko, I. A. Chuprov, Ya. N. Khassan, E. N. Kazakov, J. Gao","doi":"10.1134/S1064562424602300","DOIUrl":null,"url":null,"abstract":"<p>The accuracy of solving partial differential equations using physics-informed neural networks (PINNs) significantly depends on their architecture and the choice of hyperparameters. However, manually searching for the optimal configuration can be difficult due to the high computational complexity. In this paper, we propose an approach for optimizing the PINN architecture using a differential evolution algorithm. We focus on optimizing over a small number of training epochs, which allows us to consider a wider range of configurations while reducing the computational cost. The number of epochs is chosen such that the accuracy of the model at the initial stage correlates with its accuracy after full training, which significantly speeds up the optimization process. To improve efficiency, we also apply a surrogate model based on a Gaussian process, which reduces the number of required PINN trainings. The paper presents the results of optimizing PINN architectures for solving various partial differential equations and offers recommendations for improving their performance.</p>","PeriodicalId":531,"journal":{"name":"Doklady Mathematics","volume":"110 1 supplement","pages":"S8 - S14"},"PeriodicalIF":0.5000,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Doklady Mathematics","FirstCategoryId":"100","ListUrlMain":"https://link.springer.com/article/10.1134/S1064562424602300","RegionNum":4,"RegionCategory":"数学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"MATHEMATICS","Score":null,"Total":0}
引用次数: 0
Abstract
The accuracy of solving partial differential equations using physics-informed neural networks (PINNs) significantly depends on their architecture and the choice of hyperparameters. However, manually searching for the optimal configuration can be difficult due to the high computational complexity. In this paper, we propose an approach for optimizing the PINN architecture using a differential evolution algorithm. We focus on optimizing over a small number of training epochs, which allows us to consider a wider range of configurations while reducing the computational cost. The number of epochs is chosen such that the accuracy of the model at the initial stage correlates with its accuracy after full training, which significantly speeds up the optimization process. To improve efficiency, we also apply a surrogate model based on a Gaussian process, which reduces the number of required PINN trainings. The paper presents the results of optimizing PINN architectures for solving various partial differential equations and offers recommendations for improving their performance.
期刊介绍:
Doklady Mathematics is a journal of the Presidium of the Russian Academy of Sciences. It contains English translations of papers published in Doklady Akademii Nauk (Proceedings of the Russian Academy of Sciences), which was founded in 1933 and is published 36 times a year. Doklady Mathematics includes the materials from the following areas: mathematics, mathematical physics, computer science, control theory, and computers. It publishes brief scientific reports on previously unpublished significant new research in mathematics and its applications. The main contributors to the journal are Members of the RAS, Corresponding Members of the RAS, and scientists from the former Soviet Union and other foreign countries. Among the contributors are the outstanding Russian mathematicians.