Source:Ā Roberto Michel / Shutterstock.com
Researchers at Department of Energyās Argonne National Laboratory report that they have used artificial intelligence (AI) to dramatically reduce the time it takes to process data coming from the Laser Interferometer Gravitational-Wave Observatory (LIGO).
As LIGO and its international partners continue to upgrade their detectorsā sensitivity to gravitational waves, they will be able to probe a larger volume of the universe, thereby making the detection of gravitational wave sources a daily occurrence. This discovery deluge will launch the era of precision astronomy that takes into consideration extrasolar messenger phenomena, includingĀ electromagnetic radiation,Ā gravitational waves,Ā neutrinos, andĀ cosmic rays.
Recently, computational scientist and lead for translational AI, Eliu Huerta of the U.S. Department of Energyās (DOE) Argonne National Laboratory, in conjunction with collaborators from Argonne, the University of Chicago, the University of Illinois at Urbana-Champaign,Ā NVIDIAĀ andĀ IBM, developed a newĀ production-scaleĀ AIĀ framework that allows for accelerated, scalable and reproducible detection ofĀ gravitational waves.
This new framework indicates thatĀ AIĀ models could be as sensitive as traditional template matching algorithms, but orders of magnitude faster. Furthermore, theseĀ AIĀ algorithms would only require an inexpensive graphics processing unit (GPU), like those found in video gaming systems, to process advancedĀ LIGOĀ data faster than real time.
TheĀ AIĀ ensemble used for this study processed an entire month (AugustĀ 2017)Ā of advancedĀ LIGOĀ dataĀ in less than seven minutes, distributing the dataset overĀ 64Ā NVIDIAĀ V100Ā GPUs. TheĀ AIĀ ensemble used byĀ the team for this analysis identified all four binary black hole mergers previously identified in that dataset,Ā and reported no misclassifications.
āAs a computer scientist, whatās exciting to me about this projectĀ is that it shows how, with the right tools,Ā AIĀ methods can be integrated naturally into the workflows of scientists ā allowing them to do their work faster and better ā augmenting, not replacing, human intelligence,ā said Ian Foster, director of Argonneās Data Science and Learning (DSL) division.
Bringing disparate resources to bear, this interdisciplinary and multi-institutional team of collaborators hasĀ published a paper in Nature AstronomyĀ showcasing a data-driven approach that combines the teamās collective supercomputing resources to enable reproducible, accelerated, AI-driven gravitational wave detection.
āIn this study, weāve used the combined power ofĀ AIĀ and supercomputing to help solve timely and relevantĀ big-data experiments. We are now makingĀ AIĀ studies fully reproducible, not merely ascertaining whetherĀ AIĀ may provide a novel solution to grand challenges,ā Huerta said.
Building upon the interdisciplinary nature of this project, the team looks forward to new applications of thisĀ data-driven framework beyond big-data challenges in physics.
āThis work highlights the significant value of data infrastructure to the scientific community,ā said BenĀ Blaiszik, a research scientist at Argonne and the University of Chicago.Ā āāThe long-term investments that have been made byĀ DOE, the National Science Foundation (NSF), the National Institutes of Standards and Technology, and others have created a set of building blocks. It is possible for us to bring these building blocks together in new and exciting ways to scale this analysis and to help deliver these capabilities to others in the future.ā
Huerta and his research team developed their new framework through the support of theĀ NSF, Argonneās Laboratory Directed Research and Development (LDRD) program andĀ DOEās Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program.
āTheseĀ NSFĀ investments contain original, innovative ideas that hold significant promise of transforming the way scientific data arriving in fast streams are processed. The planned activities are bringingĀ accelerated and heterogeneousĀ computing technology to many scientific communities of practice,ā said Manish Parashar, director of the Office of Advanced Cyberinfrastructure atĀ NSF.
The new framework builds off of aĀ frameworkĀ originallyĀ proposedĀ by Huerta and his colleagues inĀ 2017. The team further advanced their use ofĀ AIĀ for astrophysics research by leveraging Argonne supercomputing resources through a two-year award from the Argonne Leadership Computing Facilityās (ALCF) Data Science Program. This led to the teamās currentĀ INCITEĀ project on the Summit supercomputer at the Oak Ridge Leadership Computing Facility (OLCF). TheĀ ALCFĀ andĀ OLCFĀ areĀ DOEĀ Office of Science User Facilities.