Applying Machine-Learning Methods to Laser Acceleration of Protons - Lessons Learned From Synthetic Data

Aug 2, 2026

Upload by Riley

Study compares three machine-learning approaches—a three-hidden-layer neural network, support vector regression, and Gaussian process regression—for learning from a synthetic dataset in the Target Normal Sheath Acceleration regime for proton acceleration. The synthetic data are generated by modifying a previously published theoretical model. After training, the models support maximizing peak proton energy and shaping proton energy spectra via laser configuration. Evaluation covers both predictive accuracy and performance on a GPU, including memory usage, with support vector regression showing strong results despite its relative simplicity.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages13

Reading Time33 min

AccessFree to download

Introduction

Machine learning for ultra-intense laser proton acceleration

Background and related work

Air Force Institute of Technology AFIT Scholar Faculty Publications 11-22-2024 Applying Machine‐Learning Methods to Laser Acceleration of Protons: Lessons Learned From Synthetic Data Ronak Desai The Ohio State University Thomas Zhang The Ohio State University J. J. Felice The Ohio State University Ricky Oropeza The Ohio State University Joseph R. Smith Marietta College See next page for additional authors Follow this and additional works at: [https://scholar.afit.edu/facpub](https://scholar.afit.edu/facpub) Part of the Computer Sciences Commons, Engineering Physics Commons, and the Plasma and Beam Physics Commons Recommended Citation Desai, R., Zhang, T., Felice, J. J., Oropeza, R., Smith, J. R., Kryshchenko, A., Orban, C., Dexter, M. L., & Patnaik, A. K. (2024) . Applying Machine‐Learning Methods to Laser Acceleration of Protons: Lessons Learned From Synthetic Data. Contributions to Plasma Physics, e202400080 . [https://doi.org/10.1002/](https://doi.org/10.1002/)[ ](https://doi.org/10.1002/)ctpp.202400080 This Article is brought to you for free and open access by AFIT Scholar. It has been accepted for inclusion in Faculty Publications by an authorized administrator of AFIT Scholar. For more information, please contact [AFIT.ENWL.Repository@us.af.mil](AFIT.ENWL.Repository@us.af.mil). Authors Ronak Desai, Thomas Zhang, J. J. Felice, Ricky Oropeza, Joseph R. Smith, Alona Kryshchenko, Chris Orban, Michael L. Dexter, and Anil K. Patnaik This article is available at AFIT Scholar: [https://scholar.afit.edu/facpub/1597](https://scholar.afit.edu/facpub/1597) Contributions to Plasma Physics ORIGINAL ARTICLE Applying Machine-Learning Methods to Laser Acceleration of Protons: Lessons Learned From Synthetic Data Ronak Desai1 | Thomas Zhang1 | John J. Felice1 | Ricky Oropeza1 | Joseph R. Smith2 | Alona Kryshchenko3 | Chris Orban1 | Michael L. Dexter4 | Anil K. Patnaik4 1 Department of Physics, The Ohio State University, Columbus, Ohio, USA | 2 Department of Physics, Marietta College, Marietta, Ohio, USA | 3 Department of Mathematics, California State University Channel Islands, Camarillo, California, USA | 4 Department of Engineering Physics, Air Force Institute of Technology, Wright-Patterson AFB, Ohio, USA Correspondence: Ronak Desai ([desai.458@osu.edu](desai.458@osu.edu)) Received: 16 July 2024 | Revised: 4 September 2024 | Accepted: 16 October 2024 Funding: This work was supported by the National Science Foundation (2109222), U.S. Department of Energy (89243021SSC000084), and Air Force Office of Scientific Research (23AFCOR004) . Keywords: laser-driven ion acceleration | machine learning | normal sheath acceleration | optimization | target ABSTRACT In this study, we consider three different machine-learning methods—a three-hidden-layer neural network, support vector regression, and Gaussian process regression—and compare how well they can learn from a synthetic data set for proton acceleration in the Target Normal Sheath Acceleration regime. The synthetic data set was generated from a previously published theoretical model by Fuchs et al. 2005 that we modified. Once trained, these machine-learning methods can assist with efforts to maximize the peak proton energy, or with the more general problem of configuring the laser system to produce a proton energy spectrum with desired characteristics. In our study, we focus on both the accuracy of the machine-learning methods and the performance on one GPU including memory consumption. Although it is arguably the least sophisticated machine-learning model we considered, support vector regression performed very well in our tests. 1 | Introduction The field of ultra-intense laser science is increasingly beginning to embrace machine learning (ML) and Bayesian optimization methods [1, 2] . This is especially true as the repetition rates of ultra-intense laser systems increase and data acquisition systems improve (e.g., Heuer et

FAQs

Which machine-learning methods are compared in the study?

The study compares a three-hidden-layer neural network, support vector regression, and Gaussian process regression.

What data are used to train the models?

Models are trained using a synthetic dataset generated from a previously published theoretical model, which is modified by the authors.

What evaluation aspects are emphasized beyond learning performance?

The paper evaluates both accuracy and computational performance on one GPU, including memory consumption.

Aug 2, 2026

Upload by Riley