Foundational Methods for Foundation Models in Scientific Machine Learning - Lecture 6

Explore the potential of foundation models for scientific machine learning in this 44-minute lecture by M. Mahoney at the University of Michigan's MICDE. Delve into the application of the "pre-train and fine-tune" paradigm, widely used in computer vision and natural language processing, to scientific computing problems. Examine the challenges and failure modes that arise when integrating data-driven machine learning methodologies with domain-driven scientific computing approaches. Learn about ongoing research to develop novel methods addressing these challenges and their large-scale implementations. Gain insights into the path towards building robust and reliable scientific machine learning models with millions to trillions of parameters, potentially revolutionizing how we approach complex scientific problems across various domains.