Description: Graph neural networks provide a valuable tool for learning relationships within structured data. Unlike convolutional neural networks, graph neural networks provide rotationally invariant interpretations of data and enable connectivity beyond regular Cartesian grids. Graphs are composed of nodes, edges, and global quantities, and can be used to learn any of these quantities. This course will cover the basics of graphs and graph convolution operations, a survey of applications in the literature, and a lab demonstrating hands-on examples using Spektral, a graph neural network package.

CUI