Graph ML: Practical Approaches for Unattributed Networks

Bio: Anwar Said is a Research Scientist at the Institute for Software Integrated Systems, Vanderbilt University. Prior to that, he was a postdoc at Vanderbilt University. He received his PhD in Computer Science from Information Technology University (ITU Lahore) under the supervision of Dr. Saeed Ul Hassan and was a member of Artificial Intelligence Lab. His research primarily intersects the fields of data science and machine learning, with a focus on graph representation learning, social network analysis, and graph theory with applications in social and molecular networks, networked control systems and healthcare. He actively participates in top conferences and journals in these domains, contributing through both publications and serving as a PC member and a reviewer. His works have been published in renowned venues such as NeurIPS, Neurocomputing, Applied Soft Computing, and TKDE, among others.

Abstract: Graph-structured data is increasingly essential across fields, from telecommunication networks to molecular chemistry, capturing complex relationships that standard ML approaches often overlook. Integrating relational patterns into deep learning models is key to building systems capable of learning, reasoning, and generalizing effectively on this type of data. However, working with graphs requires specialized techniques distinct from traditional ML workflows.

This hands-on tutorial offers an in-depth exploration of graph machine learning (Graph ML) approaches, covering foundational techniques for representing and preparing graph-structured data. Participants will learn methods and tools for handling graph data and explore several applications. In an unattributed social network setting, we will introduce PropEnc, a feature engineering approach designed to construct expressive node features for the graph classification task. The session culminates in implementing a graph neural network pipeline for training on a social network dataset. No prior experience with Graph ML is required—this session provides a foundation in Graph ML, equipping participants with practical skills to implement graph ML pipelines.