Project Grant 2310679

The Trustees of the University of Pennsylvania received a $275,000 project grant from the National Science Foundation's Mathematical and Physical Sciences program (CFDA 47.049) to conduct research on geometrization approaches toward understanding deep learning. Specifically, the three-year award funds research projects analyzing symmetries in trained deep neural networks, examining dynamics of deep learning training, and investigating how deep learning separates data across neural network layers. The goal is to develop a comprehensive statistical and mathematical understanding of deep learning that demystifies these complex models and makes them more interpretable and efficient for scientific and engineering applications. The research also aims to boost professional development for statisticians and data scientists. Completion of the projects is expected to result in a geometrization of deep learning methodologies. The award furthers the NSF's mission to strengthen the nation's scientific enterprise through advancing mathematical and physical sciences.