Inaugural Scientific Machine Learning Workshop Looks Toward the Future

Nearly 100 attendees participated in the inaugural workshop on Scientific Machine Learning (SciML), held April 3 and 4 at the Oden Institute for Computational Engineering and Sciences at The University of Texas at Austin. The event hosted by Rachel Ward and Tan Bui-Thanh, co-directors of the Center for Scientific Machine Learning, featured 21 speakers and created an environment for fostering collaboration, establishing central challenges, and point to directions  in current and future research.

“Speakers were invited from domestic and international academia (computational math, computational science and engineering, computer sciences, engineering) and industries. The goal was to have interactions from these fields/disciplines to provide an overview of the state-of-the-art methods, current research challenges and the actual need from industries,” stated Bui-Thanh.

While many people outside of the computational scientific community probably don’t give SciML much thought, this rapidly emerging field is the computational modeling architecture that underpins research in areas that affect everyday life and is at the intersection of both pure and applied mathematics.

“Discovery in this field is important for the broader STEM communities. We, and our peers, have developed advanced SciML approaches for solving practical problems such as earthquakes, storms and hurricane, medical imaging, climate, astrophysics, radar, fusion energies, dense applications, etc. SciML could have an important impact on society, technologies, economy, and national security,” said Bui-Thanh.

The goal of the workshop, which included students in the Computational Science Engineering and Math (CSEM) program at the Oden Institute, was to provide a peer-to-peer forum helping put research in perspective and for the exchange of ideas. Speakers included faculty within The Oden Institute, and experts from academic institutions across the globe, as as well as representatives from a number of industries including Ata Engineering, Exxon Mobile, MathSci.ai, Raytheon, and Siemens.

SciML is growing exponentially fast. Bui-Thanh said regular meetings are necessary for exchanging ideas to make progress in sciences. “We established the first workshop of its kind and we hope it will be an annually rotating event between different organizations and universities,” he said.

“Overall, we were pleased with this inaugural workshop. Attendees commented to us that they were very impressed with the high quality of talks and breadth of topics covered in scientific machine learning,” said Ward.

Oden CSEM students and research assistants found the event beneficial. Research assistant Wesley Lao, who works with Bui-Thanh, enjoyed the different perspectives from experts in the field. “I think their insights into how machine learning can be applied to physics-based problems will inform how I approach my work for the near future.” Lao has been admitted to the CSEM Ph.D program.

Second year Ph.D student Keith Poletti said the workshop introduced him to novel methods in Operator Learning. “Multiple talks focused on applying these methods to specific problems and the challenges faced. Soledad Villar, who is a student of Rachel Ward, presented a method to enforce conservation laws/symmetries. She had a rigorous approach and left me curious to learn more.”

Workshop attendee Hai Nguyen, who works with Bui-Thanh said the workshop was an excellent forum for staying updated on the latest industrial and academic trends in scientific computing and machine learning. “By participating in the conference I was able to learn a plethora of new information and broaden my knowledge in this dynamic field. I hope the conference continues to be held annually.”

The Center for Scientific Machine Learning at the Oden Institute is focused on the opportunities and challenges of machine learning in the context of complex applications across science, engineering, and medicine. The greatest challenges facing society — clean energy, climate change, sustainable urban infrastructure, access to clean water, personalized medicine and more — by their very nature require predictions that go well beyond the available data. Scientific machine learning achieves this by incorporating the predictive power, interpretability and domain knowledge of physics-based models.