Jiong Zhu receives 2024 CSE David J. Kuck Dissertation Prize

Recent alum Jiong Zhu (CSE PhD 2024) has been selected as the 2024 recipient of the David J. Kuck Dissertation Prize in recognition of his outstanding dissertation on graph neural networks for complex data. Awarded to one individual annually, the Kuck Dissertation Prize aims to recognize exceptionally impactful dissertations by CSE PhD students. 

Zhu’s dissertation, titled “Advancing Graph Neural Networks for Complex Data: A Perspective Beyond Homophily,” explores the capabilities and limitations of graph neural networks (GNNs) and introduces novel methodologies to improve their accuracy, robustness, fairness, and scalability. GNNs are a type of artificial intelligence that extend the principles of machine learning to graph-structured data—data that can be represented as networks of interconnected nodes and edges. This makes GNNs particularly useful for applications like social networks, fraud detection, and bioinformatics.

Most GNNs improve performance by relying on homophily, the principle that connected nodes tend to be similar. Zhu’s research addresses the shortcomings of GNNs in heterophilous scenarios, where connected nodes are dissimilar. His work reveals that by understanding and leveraging different patterns of node connections beyond traditional homophily, it’s possible to significantly enhance GNN performance. Zhu’s research not only improves accuracy but also enhances robustness against adversarial attacks, ensures fairness, and streamlines distributed training on massive datasets.

Zhu’s work has received widespread academic recognition, including over 1,450 citations, and has been presented at top conferences such as NeurIPS, AAAI, and KDD. His dissertation’s impact on the field has been underscored by invited talks in academia and industry, highlighting both the theoretical significance and practical relevance of his contributions.

Zhu is now a full-time Applied Scientist at Amazon, where he is working on building foundational large language models (LLMs) for Amazon’s storefront.

“Jiong’s dissertation on graph neural networks has undoubtedly advanced the field, offering new insights and methodologies that will shape future research and applications,” said Zhu’s advisor Professor Danai Koutra. “I have no doubt that he will continue to make significant contributions to the field in his future endeavors.”