Machine learning takes root in nuclear physics


Credit: Pixabay/CC0 Public Domain

Scientists are beginning to turn to the new tools provided by machine learning to help save time and money. In the past several years, nuclear physics has seen a series of machine learning projects brought online, with several research papers published on the topic. Now, 18 authors from 11 institutions are summarizing this explosion in action with the help of artificial intelligence in “Machine Learning in Nuclear Physics,” research recently published in Modern Physics Assessments.

“It was important to document the work that was done. We really want to raise the level of use machine learning in nuclear physics To help people see the breadth of activities,” said Amber Bunlin, lead author of the paper and co-director of computational sciences and technology at the US Department of Energy’s Thomas Jefferson National Accelerator Facility.

As the paper collects and summarizes the major work in this area to date, Boehnlein hopes that it will serve as a educational resources For interested readers, as well as a roadmap for future endeavors.

“It provides a benchmark that people can use as they progress to the next stage,” she said.

A revolution in machine learning

After attending a workshop on Exploring Artificial Intelligence at Jefferson Lab in March 2020 and publishing a follow-up report, Bunlin and two of her co-authors, Wittold Nazarevich and Michael Kucera, were inspired to move forward. With 15 colleagues representing all subfields of nuclear physics, they decided to survey the state of machine learning projects in nuclear physics.

They started at the beginning. As the authors describe, the first significant work using machine learning in nuclear physics used computer experiments to study nuclear properties, such as atomic masses, in 1992. Although this work hinted at the potential of machine learning, its use in this field has remained scanty for most of two decades. In the past several years, that has changed.

Machine learning, which involves building models that can perform tasks without explicit instructions, requires computers to do specific things, including complex mathematical operations. With recent advances, computers can better meet these requirements, allowing physicists to more easily integrate machine learning into their work.

“This paper could have been less interesting in 2019, because there wasn’t enough work to be indexed,” Bunlin said. “But now, there is a lot of work to cite due to the increased use of technologies.”

Today, machine learning encompasses all scales and ranges of research power, from investigations into the basic building blocks of matter to inquiries into the life cycles of stars. They are also found in the four subfields of nuclear physics: theory, experiment, accelerometer, operations, and data science.

said co-author Kuchera, an associate professor of physics and computer science at Davidson College.

Machine learning models can be used to help design and carry out experiments in nuclear physics. They can also be used to help analyze the data of those experiments, which are often in excess of petabytes.

“I expect machine learning to become an integral part of data collection and analysis,” Kuchera said.

Machine learning will speed up these processes, which could mean that less time and money will be required for radiation time, computer use, and other experimental costs.

Connecting theory and experiment

However, machine learning has so far developed the strongest foothold in nuclear theory. Nazarewicz, a nuclear theorist and chief scientist at the Rare Isotope Package Facility at Michigan State University, is particularly interested in this topic. He says machine learning can help theorists make advanced calculations faster, improve and simplify models and predictions, and help theorists understand the uncertainties in their predictions. It can also be used to study phenomena that researchers cannot experiment with, such as supernova explosions or neutron stars.

“Neutron stars are not easy to use,” Nazarewicz said.

It uses machine learning to study nuclei and very heavy elements, which have too many protons and neutrons in their nuclei to be observed experimentally.

“I find the results to be the most impressive in the theoretical community, particularly the low-energy theory community with which Wittold is associated,” said Bonlain. “They seem to really embrace these technologies.”

Boehnlein said theorists are also beginning to adopt these techniques at Jefferson Lab in their studies of protons and neutron structures. Specifically, machine learning can help extract information from complex theories, such as Quantum chromodynamicsThe theory that describes the interactions between quarks and gluons that make up protons and neutrons.

The authors anticipate that the involvement of machine learning in both theory and experiment will independently accelerate these subfields, as well as better connect them to accelerate the whole loop of the scientific process.

“Nuclear physics helps us make discoveries to better understand the nature of our universe, and it is also used in societal applications,” Nazarewicz said. “The faster we can make the cycle between experiment and theory, the faster we can get to discoveries and applications.”

As machine learning continues to grow in this field, the authors expect to see more developments and broader applications integrating this tool.

“I think we are only in the infancy of applying machine learning to nuclear physics,” Boonelain said.

And along the way, this paper will serve as a reference, even to its authors.

“I hope that the paper will be used as a resource for understanding the current state of machine learning research, allowing us to build from these efforts,” Kuchera said. “My research focuses on machine learning methods, so I will use this paper as a window into the state of machine learning via nuclear power. Physics Immediately.”

The theory suggests that quantum computers should be significantly faster at some learning tasks than classical machines

more information:
Amber Boehnlein et al, Symposium: Machine Learning in Nuclear Physics, Modern Physics Assessments (2022). DOI: 10.1103/RevModPhys.94.031003 . the paper Also available on arXiv.

the quote: Machine Learning Takes Root in Nuclear Physics (2022, October 13) Retrieved October 13, 2022 from

This document is subject to copyright. Notwithstanding any fair dealing for the purpose of private study or research, no part may be reproduced without written permission. The content is provided for informational purposes only.

Leave a Comment