Physics and Computational Experts Help Raise $ 15 Million to Advance AI and Data Analytics
As scientific datasets get larger and larger, data processing algorithms become more complex. Artificial intelligence (AI) has become a solution to efficiently analyze these massive data sets, and new types of computer processors, such as graphics processing units (GPUs) and field programmable gate arrays (FPGAs), help speed up the work of AI algorithms. . This combination of AI and new types of processors is leading to a revolution in the field of data analysis.
In an effort to change direction in the application of real-time AI at scale, the National Science Foundation (NSF) has funded $ 15 million to support the advancement of scientific knowledge and discovery with the institute Accelerated AI Algorithms for Data-Driven Discovery (A3D3). . Its mission is to integrate AI algorithms into new processors to support analyzes of these unprecedented data sets.
âAI-assisted analysis of multidisciplinary datasets will be essential in helping researchers locate and explore trends that may lead to new discoveries,â said UC San Diego Chancellor Pradeep K. Khosla. âThe new multidisciplinary and geographically distributed A3D3 institute, supported by NSF’s Harnessing the Data Revolution (HDR) program, will lead the way with a collaborative team of researchers from UC San Diego, Caltech, Duke University, MIT, Purdue University , UIUC, University of Minnesota, University of Washington and University of Wisconsin-Madison.
Influence of UC San Diego
Javier Duarte of UC San Diego, an assistant professor in the physics department that collaborates with researchers at the San Diego Supercomputer Center (SDSC), will serve as the institute’s principal investigator (PI) and a representative of the institute’s council of the university, as well as the equity and career representative on the board. In this role, he will also co-supervise the post-baccalaureate program with Mia Liu at Purdue University. Frank WÃ¼rthwein, Acting Director of the SDSC, will attend, as will Amit Majumdar, who heads the Data Enabled Scientific Computing division of the SDSC. Additionally, UC San Diego postdoctoral researcher Daniel Diaz and graduate students Raghav Kansal, Farouk Mokhtar, and Anthony Aportela will develop accelerated AI algorithms.
Duarte said his work would be split between developing ultra-fast machine learning algorithms deployed in specialized hardware, such as FPGAs, which can be used to process sensor data in real time, and developing computer pipelines. heterogeneous to allow faster processing of large scientific data.
Target scientific fields for rapid AI and dissemination
To take full advantage of rapid AI, the A3D3 Institute will target fundamental problems in three scientific fields: high energy physics, multi-messenger astrophysics and systems neuroscience.
âA3D3 is working closely in these areas to develop custom AI solutions to process large data sets in real time, dramatically improving their discovery potential,â said Duarte. “The ultimate goal of A3D3 is to build the institutional knowledge essential for real-time applications of AI in any scientific field.”
Duarte also noted that the A3D3 will equip scientists with new tools to deal with the coming data deluge through dedicated outreach efforts.
âThe post-baccalaureate program, for example, will specifically aim to help underrepresented minority students, identifying as Black, Latinx, Indigenous, female or LGBT +, from institutions without extensive research opportunities, to gain experience of valuable research to ‘bridge the gap’ between undergraduate and graduate programs, âhe said.
The Director and PI of the A3D3 Institute is Shih-Chieh Hsu of the University of Washington, a colleague and former student from WÃ¼rthwein, who gave an example of the potential impact of work at the institute.
âAt the Large Hadron Collider (LHC), the data processing challenge is enormous. With future aggregate data rates exceeding one petabit per second, data rates at the LHC exceed all other devices in the world, âexplained Hsu. âThe objective of A3D3 is to build a series of tools that will enable all this information to be processed in real time using AI. Through the use of AI, A3D3 aims to perform advanced analyzes, such as anomaly detection and particle reconstruction on all collisions occurring 40 million times per second! “
Real-time analyzes in astrophysics and neuroscience
For UC San Diego-based projects outside SDSC, such as Traveler.
âVoyager is based on dedicated AI hardware from Habana, while NRP includes both more conventional FPGAs and GPUs,â noted WÃ¼rthwein.
Duarte noted that in the field of multi-messenger astrophysics, A3D3 will work to integrate AI to rapidly and computer-process data from telescopes, neutrino detectors and gravitational wave detectors efficiently to quickly identify the astronomical events corresponding to the most violent. phenomena in the cosmos.
“The ability to identify and further distribute these events as astronomical alerts enables the entire transient astronomical community to correlate observations and understand astrophysical phenomena through several different forces,” said Duarte.
Amy Orsborn, an assistant professor in the Department of Electrical and Computer Engineering and the Department of Bioengineering at the University of Washington, explained that in systems neuroscience, A3D3 strives to uncover the computations that neural networks at the scale of the brain perform to process sensory and motor functions. information during the behavior. To do this, A3D3 will develop and implement high throughput, low latency AI algorithms to process, organize and analyze massive neural data sets in real time.
âThese real-time analyzes will allow new approaches to probe brain function, such as closed-loop causal manipulations. Applying powerful AI methods to systems neuroscience will dramatically advance our ability to analyze and interpret neural activity and its relationship to behavior, âsaid Orsborn.
According to the NSF, the five new institutes will enable breakthroughs through collaborative and co-designed programs to formulate innovative, data-intensive approaches to address critical national challenges. The first results are expected by 2023.
This project is supported by the NSF (grant # 2117997).