3.0

credits

Average Course Rating

A course on data science methods for graphs. Topics encompass graph signal processing, including the graph Fourier transform, graph signal sampling and convolutional graph neural networks (GNNs); graphon signal processing, graphon neural networks and convergence and transferability analyses of GNNs; and modern graph deep learning methods, including more efficient GNN architectures and training algorithms (e.g., gradient sampling and computational sampling) and graph dataset distillation. A mix of theory and application, the course includes labs and/or a final project in PyTorch. PyTorch knowledge is not required, but students must be familiar with Python.