February 17
@
12:00 PM
–
1:00 PM
Harnessing Geometry and Topology in Machine Learning for Biomedical Discovery
Dhananjay Bhaskar, PhD
Postdoctoral Researcher, Yale University
Yale – Boehringer Ingelheim Fellow
Kavli Institute for Neuroscience Fellow
Visiting Scholar in Engineering, Brown University
Abstract:
Understanding the shape of data through geometry and topology offers a powerful perspective with the potential to drive breakthroughs in biomedical science and engineering. This talk explores how these mathematical frameworks, integrated with machine learning, can drive innovation in drug discovery, understanding cellular communication, and unraveling the dynamics of brain activity.
I will introduce learnable geometric scattering, a novel machine learning framework that overcomes the key limitations of conventional graph neural networks. By capturing nuanced structural features, this approach opens new frontiers in drug discovery and protein conformational analysis. Building on this, I will demonstrate how combining geometric scattering with topological data analysis uncovers hidden patterns in cell signaling and neural activity, shedding light on wound healing mechanisms and schizophrenia diagnosis. Finally, I will highlight how these tools can deepen our understanding of learning in artificial neural networks, paving the way for the development of biologically inspired, more expressive, and robust machine learning models.
This talk will emphasize the practical applications of these approaches and their transformative potential for addressing real-world challenges in health and medicine.
Print PDF