Computational Mathematics & Statistics Seminar by Jia He: Machine Learning on Graphs: Graph Signal Processing and Graph Substructure Learning

Speaker:

Jia He, PHD, Illinois Institute of Technology

Title:

Machine Learning on Graphs: Graph Signal Processing and Graph Substructure Learning

Abstract:

Machine learning on graphs is an interesting and challenging task with broad applications. Many applications require incorporating information about the link structure of the graph as well as the features on the nodes and edges into the machine learning model. In this talk we distinguish two types of graph learning problems: graph signal processing and graph substructure learning. The first focuses on learning the signals supported on graphs, i.e., the graph-structured data. We present two graph convolutional neural network models that are used to infer power line outage from measurements. The second focuses on learning graph structures and substructures. We present the well-known neural network models and our recent work on this topic, and compare their expressive power.

 

Computational Mathematics & Statistics