Physics-informed Machine Learning: The Next Evolution in Neural Network Development

doi: 10.53962/hn7c-20eb

Originally published on 2023-03-26 under a CC BY 4.0

Authors

Summary

Machine Learning (ML) is one of the fastest-growing subsets in artificial intelligence. One of the persistent challenges to ML model development is data quality and availability. Many researchers across a range of disciplines have begun leveraging Partial Differential Equations (PDEs) in the physical, biological, and other hard sciences to answer otherwise intractable problems resulting from the lack of data availability and issues with data quality. This approach is appropriate. While PDEs are not the only answer to all binary or multi-classification problems in ML modeling, they provide a series of physics-inspired ML modeling methods capable of addressing a set of challenges that were previously difficult, if not impossible, to infer. This presentation will focus on PDE utility.

Main file

Physics-Informed Machine Learning Tech Talk.pptx