Yao Ma 馬耀

Assistant professor in the Department of Computer Science at New Jersey Institute of Technology (NJIT)

 

Biography:

Yao Ma is an assistant professor in the Department of Computer Science at New Jersey Institute of Technology (NJIT). He received his Ph.D. in Computer Science from Michigan State University (MSU) in 2021. His major research interest lies in Graph Neural Networks (GNNs) for representation learning on graph-structured data. He has significantly contributed to the fundamental research and practical research of GNNs, which leads to numerous innovative works in top-tier conferences such as KDD, WWW, SIGIR, WSDM, ICDM, ICML, NeurIPS, and CLR. He was the leading organizer and presenter of two well-received tutorials on GNNs at AAAI’ 2020, AAAI’ 2021, KDD’ 2020, and KDD’ 2021, attracting more than 1000 attendees intotal. His recent book Deep Learning on Graphs has attracted tens of thousands of downloads from more than 100 countries. He received the Outstanding Graduate Student Award (2019-2020) from the College of Engineering at MSU.

 

 

 

Session Title: A Unified View on Graph Neural Networks as Graph Signal Denoising

Abstract:

Graph structured data are ubiquitous in the real world such as social networks, molecular graphs, and emerging among a plethora of other diverse domains. Therefore, it is of great research importance to design advanced algorithms for representation learning on graph-structured data to facilitate improved predictions across numerous computational methods. Graph Neural Networks (GNNs), which generalize the deep neural network models to graph-structured data, pave a new way to learn representations effectively. This talk presents one of the key operations of GNNs, i.e., graph filtering operation. Specifically, I will present some commonly used graph filtering operations and then demonstrates how they can be observed from a unified graph signal denoising perspective. Such a unified view across GNNs not only provides a new perspective to understand a variety of graph filtering operations but also enables us to develop a unified graph neural network framework. This unified framework further paves a new way to conveniently design novel GNNs tailored for downstream tasks with various requirements