Geometric Deep Learning For Molecular Modeling
Program for Mathematical Genomics, Columbia University
Offered: 2022
Deep learning is having a significant impact on modeling molecular systems, offering valuable applications in quantum chemistry, drug discovery, and structural biology. Like physical theories, neural network architectures modeling physical and biological systems must obey the symmetries of the phenomena they represent. This principle is known as equivariance, and it constrains the design of neural networks.
In this class, we will explore the mathematical tools essential for geometric deep learning, such as group theory and representation theory. We will then discuss how these tools are used to construct group equivariant neural networks. Additionally, we will examine the applications of geometric deep learning in cheminformatics and protein biology. Finally, we will provide an in-depth analysis of the geometric aspects of AlphaFold2.
Symmetries and Group Theory for Computational Biologists
Harvard Medical School
Offered: 2019
Aspects of Quantum Field Theory
Part III of the Mathematical Tripos, Cambridge University
Offered: 2012-2014