Knowledge-Guided Machine Learning Workshop Brings Researchers to Campus

CSE DSI hosted a one-day workshop on August 7, 2024, on Knowledge Guided Machine Learning (KGML2024): A Framework for Accelerating Scientific Discovery. The workshop was organized by Vipin Kumar, CSE DSI Director and Regents Professor in the Department of Computer Science & Engineering, and Anuj Karpatne and Xiaowei Jia, both CS&E alumni and now faculty at Virginia Tech and the University of Pittsburgh, respectively. Also involved were members of the CSE DSI leadership team, including Drs. Shashi Shekhar and  Michael Steinbach.

KGML is an emerging field of research where scientific knowledge is deeply integrated into ML frameworks to produce solutions that are scientifically grounded, explainable, and likely to generalize on out-of-distribution samples, even with limited training data. The major goal of this workshop was to explore the depth and diversity of research methodologies being explored in KGML for a wide range of scientific applications. 

The day began with an introductory tutorial on KGML by Anuj Karpatne, followed by a talk on knowledge-guided foundation models by Xiaowei Jia. Several talks addressed the use of KGML in a variety of disciplines, including hydrology, Biodiversity science, and physics.  Two panels discussed “Navigating the KGML Landscape: Challenges and Opportunities” and “KGML in the Age of Generative AI.” 

These discussions helped expose gaps in the current state of KGML research and identify novel opportunities for advancing AI  while accelerating discoveries in problems of high societal relevance. The animated discussions were enhanced by the fact that several participants in the NSF-sponsored Workshop on AI-Enabled Scientific Revolution held the previous day stayed on to attend KGML2024. 

All the talks were recorded and can be accessed from the KGML2024 Workshop website on the Agenda page, along with slides.

A KGML poster session preceded the workshop the evening before and featured several posters by graduate students in various departments in the College of Science and Engineering, including Computer Science & Engineering, Physics, AI-LEAF (AI institute), and Civil, Environmental, and Geo-Engineering.

The KGML2024 workshop was a continuation of previous workshops (KGML2020 and KGML2021) held as part of a project funded by the NSF's Harnessing the Data Revolution (HDR) program as well as annual workshops held at the AAAI Fall symposium series during Fall 2020, 2021, and 2022 and the KGML Bridge Program at AAAI 2024. 

“The KGML framework, pioneered here at the University of Minnesota, is seeing applications across nearly all science and engineering disciplines,” said Kumar.  “Artificial intelligence and knowledge-guided machine learning have limitless potential to address global societal challenges, given their applications to all science and engineering disciplines. These workshops [KGML2024 and the  NSF-Sponsored Workshop on AI-Enabled Scientific Revolution] and resulting reports will inform the scientific agenda at the interface of AI for decades to come.”