AMS 525, Geometric Deep Learning
In the first part of the course, we will cover programming in Python, from its basic
libraries up
to the implementation of advanced deep learning models such as CNNs, RNNs, GNNs and
Transformer networks. The practical success of many of these models in high dimensional
problems such as image processing, playing GO, or protein folding comes from the predefined
regularities in the underlying low-dimensional geometric structure of the data. Therefore
in the second part of the course, we will extend the aforementioned deep learning
models and their implementations to graphs and manifolds in spatial and spectral domains.
The focus will be on the implementation of the models in Python and their practical
applications.
Note: Instructor consent
Summer, 3 credits, ABCF grading
May be repeated for credit