Machine learning is helping scientists save time and money. Nuclear physics has seen a lot of machine learning projects come online in the last few years. The explosion of artificial intelligence-aided work in nuclear physics was summarized by 18 authors in a recent paper.
Documenting the work that has been done is important. The Department of Energy wants to raise the profile of machine learning in nuclear physics so that people can see the breadth of the activities.
The paper gathers and summarizes major work in the field so far that Boehnlein hopes it can act as an educational resource for interested readers.
She said that it provided a benchmark that people can use as they move forward.
The machine learning revolution is happening.
Boehnlein and two of her co-authors were inspired to go further after attending a workshop exploring artificial intelligence at Jefferson Lab in March 2020. A survey of the state of machine learning projects in nuclear physics was conducted by a group of 15 colleagues.
At the beginning, they began. The first significant work using machine learning in nuclear physics was done in 1992. The use of machine learning in the field remained low for more than two decades. The last few years have seen that change.
Machine learning requires computers to do certain things, including complex calculations. Physicists have been able to incorporate machine learning into their work due to recent advances in computers.
There wouldn't have been enough time to catalog this paper in 2019. There is a lot of work to be done because of the increased use of techniques.
Machine learning can encompass all scales and energy ranges of research, from investigations of matter's building blocks to inquiries into the life cycles of stars. It is found across the four subfields of nuclear physics.
A comprehensive, collective resource that bridges the efforts in our subfields, which will hopefully spark rich discussions and innovation across nuclear physics was compiled by the authors.
Experiments in nuclear physics can be done with machine learning models. They can be used to help with the analysis of the data from those experiments.
"I expect machine learning to become a part of our data collection and analysis."
Less time and money is needed for beamtime, computer usage, and other experimental costs if machine learning is speeded up.
Theory and experimentation are connected.
Machine learning has emerged as the strongest foothold in nuclear theory so far. The nuclear theorist and chief scientist at Michigan State University is interested in this topic. He claims that machine learning can help theorists do advanced calculations faster, improve and simplify models, make predictions, and understand uncertainties of their predictions. It can be used to study phenomena that can't be studied in an experiment.
Neutron stars are not easy to use.
He uses machine learning to study elements that have a lot of protons and neutrons in their nucleus.
The low energy theory community that Witold is associated with is the most impressive in the theory community. They seem to be using these techniques.
Physicists at Jefferson Lab have begun to use these techniques in their studies of protons and neutrons. Machine learning can be used to extract information from complicated theories such as quantum chromodynamics.
According to the authors, machine learning's involvement in both theory and experiment will speed up these subfields independently, and it will also better connect them to speed up the whole scientific process.
Nuclear physics helps us understand the nature of our universe and is also used for societal applications. The faster we can do the cycle between theory and experiment, the quicker we will arrive at discoveries.
The authors expect to see more developments in machine learning as it continues to grow.
The application of machine learning to nuclear physics is still in its infancy.
The paper will act as a reference for other authors.
The paper can be used to understand the current state of machine learning research. I will use this paper as a window into the state of machine learning in nuclear physics right now because my research is centered on machine learning methods.
More information: Amber Boehnlein et al, Colloquium : Machine learning in nuclear physics, Reviews of Modern Physics (2022). DOI: 10.1103/RevModPhys.94.031003 . The paper is also available on arXiv. Journal information: Reviews of Modern Physics
All Rights Reserved © 2024
