Doctoral Defense - Harry Shomer

About the Event

The Department of Computer Science & Engineering

Michigan State University

Ph.D. Dissertation Defense

June 10th, 2025 at 1:30pm EST

EB 3105 & https://msu.zoom.us/j/95213562662

Passcode: Upon Request from Vincent Mattison or Advisor

Abstract

ENHANCING LINK PREDICTION ON GRAPHS: A MULTIFACETED APPROACH

By: Harry Shomer
Advisor: Dr. Jiliang Tang

Graphs are a common way of representing real-world structured data. A graph is composed of nodes connected with one another via edges (i.e., links), where an edge models how nodes are related to one another. Due to the prevalence of graph-structured data, machine learning on graph data has exploded in popularity in the past decade.

Link prediction is a fundamental task on graphs, which attempts to predict unseen links in a graph. Link prediction has a multitude of real-world applications, including in recommender systems, knowledge graphs, and biology. In recent years, a flurry of methods have been introduced that make use of graph neural networks (GNNs) for this task. However, we find that two main limitations impede our ability to create effective link prediction models that can perform in real-world settings. First, we find that the current method of evaluating link prediction models is both unrealistic and too easy, resulting in inflated model performance that doesn't reflect real-world performance. Second, we observe that current methods are limited in their ability to model various patterns in link formation. This poses a challenge in real-world datasets where links can form for several reasons.

In this talk, I will talk about our recent research that uncovers both fundamental limitations. I will then introduce our attempts to combat these problems, through the design of a new evaluation strategy and more expressive model design. Through the introduction of these new approaches, we help better promote the use of link prediction in more realistic scenarios that can occur in the wild.