The field of nuclear science, which was once secretive, makes rapid progress by open sourcing
When you hear the word “nuclear science,” you may imagine that you are dealing with top-secret documents and files in a tightly guarded research laboratory. GitHub summarizes that various discoveries and progress have been made as a result.
Open source is fueling the future of nuclear physics GitHub
The successful
Announced that the Ministry of Energy has officially confirmed 'ignition' by achieving output exceeding energy input in nuclear fusion experiment - GIGAZINE
In the fusion experiment conducted at Lawrence Livermore National Laboratory, multiple simulations were performed using a supercomputer before experimenting with the actual device. At that time, complex physics simulations were performed using open source software released by multiple nuclear science institutes such as Idaho National Laboratory and Argonne National Laboratory . As a result, the success of nuclear fusion experiments using open source has paved the way for researchers at other nuclear science institutions studying nuclear fusion reactions.
However, from the beginning, nuclear science institutions were not in favor of open sourcing software and source code . I kept my doctrine.
In 2002, the U.S. Department of Energy issued a recommendation to ``disclose the source code as much as possible unless there is a reason not to disclose it'', but the reason was ``falsification of the code by a malicious person''. However, many research groups, including those in the field of nuclear science, did not respond to the release of the source code.
Derek Gaston , who works as a mathematician at the Idaho National Laboratory, developed a framework called MOOSE in 2008 that facilitates simulations on supercomputers. At this time, Mr. Gaston said, ``I wanted as many researchers as possible to use my invention, so I intended to make it open source from the beginning.'' Contrary to this intention, Idaho National Research The office kept MOOSE private until 2014.
Gaston and Cody Perman said they spent a lot of time in meetings explaining the benefits of open sourcing research and how tools like GitHub work. As a result, in 2014, Idaho National Laboratory released MOOSE as open source.
Clint Finley of GitHub's '
According to Finley, conventional nuclear research published in academic journals often involves experiments and analyzes using software that is not widely available, and as a result, the results of published research are often used as the basis for research. It was difficult to conduct further research and interpret the data.
Ethan Peterson of the Massachusetts Institute of Technology's Plasma Science and Fusion Center said, ``In the past, if you wanted access to the source code of software, you would send someone an email and a contract agreeing to limit the use of the source code to a specific use. I often signed papers,” he recalls. Also, April Novak of Argonne National Laboratory said, 'Depending on the size of the laboratory, it could take several years to access the source code. No. The range of things that can be researched becomes narrower and narrower.”
Although open sourcing is progressing in the field of nuclear science, not all source code is disclosed for released software. 'We do not open source code that has functions that can be diverted to nuclear weapons or weapons,' Novak said. Mr. Finlay describes the way open sourcing should be done in nuclear science institutions as 'open but not too open'.
Nevertheless, Lawrence Livermore National Laboratory released its own computational framework called ' MFEM ' in 2010, and has been an early contributor to the open source movement. MFEM is software that aims to simplify the process of writing the code necessary for scientists to perform simulations by providing discretization and parallel scalability .
In addition, Lawrence Livermore National Laboratory is also developing an approach called `` Cognitive Simulation (CoqSim) '' that combines AI and machine learning technology with MFEM.
CoqSim is a system that can learn using a huge amount of data collected from experiments, incorporate a unique model based on past experimental data, and predict the success probability of various experiments. In the confirmation of `` ignition '' in the nuclear fusion experiment conducted at Lawrence Livermore National Laboratory in 2022, CoqSim has a success probability of 17% presented in previous experiments, ``success with a probability of over 50%''. I made a much higher prediction.
In CoqSim, the data obtained from each experiment is fed back to the system and used to improve predictions of future simulations and success probabilities. It is also predicted that it will be possible to promote more innovative thinking that eliminates human preconceptions and prejudices.
Also, Lawrence Livermore National Laboratory has open sourced a tool called ' Merlin ' to simplify the management of AI workflows on supercomputers. Merlin enables researchers to efficiently manage large numbers of jobs in machine learning tasks and run millions of simulations more efficiently.
Many of these source codes and software, which have been open sourced with nuclear science in mind, can be applied in various fields using supercomputers. MFEM, for example, is used in the wider scientific community, including the heart simulation toolkit
Also, during the epidemic of the new coronavirus, researchers at Lawrence Livermore National Laboratory used Merlin to predict the occurrence of infectious diseases and modeled antibodies.
By making source code etc. open source by nuclear science institutions, anyone can participate in research regardless of scientific background and it is possible to advance science. Also, some scientists have a lack of knowledge in the field of software engineering, so open sourcing gives them the opportunity to get feedback from many experienced software engineers. ``The scientific community is trying to learn as much software expertise as possible, but sometimes they learn from professional engineers,'' Peterson said.
In addition, Mr. Peterson said, ``Reporting bugs on open source source code is part of participating in research,'' and is seeking support from many professional engineers.
Opening up repositories and working collaboratively with experts and non-experts is a far cry from the secretive image of nuclear science. However, it is a very valuable experience for scientists to get feedback from outside the research organization to which they belong, and it is an important factor in the progress of science.
``We treat all the feedback we receive as if we were submitting it to an academic journal,'' said Zanio Kolev of Lawrence Livermore National Laboratory. If you knew that it would be checked, you should be able to create more detailed and polite code.'
``Secrecy is certainly important, but in nuclear science, having a certain degree of openness is essential for scientific progress,'' Finley said.
Related Posts:
