Events

MaD Seminar: Beyond Static Memorization: Structural Inference and Generative Reasoning in Knowledge Graphs

Speaker: Joyce Whang (KAIST)

Location: 60 Fifth Avenue, Room 150

Date: Thursday, October 8, 2026

Abstract: Graph-structured data is fundamental to modern AI, driving applications from complex reasoning engines to personalized recommendation systems. At the core of these capabilities are Knowledge Graphs (KGs), which formalize real-world facts as interconnected entities and relations. While traditional KG representation learning enables the integration of this knowledge into machine learning pipelines, it relies heavily on the static memorization of fixed embeddings. This approach creates a critical bottleneck for generalizability, particularly in dynamic, open-world settings. This talk explores a paradigm shift from rote memorization toward deep structural inference. First, it introduces novel methodologies that capture the intrinsic topological relationships within KGs, enabling robust, inductive inference on entirely unseen entities and relations. Second, extending this framework to complex, hyper-relational KGs demonstrates how a purely structure-based approach achieves state-of-the-art performance on various reasoning tasks. Finally, the talk highlights the generative potential of this structural approach by proposing a novel framework that synthesizes new complex hyper-relational facts, advancing the frontier of autonomous knowledge discovery and demonstrating how structural learning effectively complements Large Language Models (LLMs).

Bio: Joyce Jiyoung Whang is an Associate Professor in the Department of AI Computing at KAIST, where she has led the Big Data Intelligence Lab since July 2020. Before joining KAIST, she was an Assistant Professor of Computer Science and Engineering at Sungkyunkwan University (SKKU) from March 2016 to June 2020. She received her Ph.D. degree in Computer Science from the University of Texas at Austin in December 2015 under the supervision of Professor Inderjit Dhillon. She serves as Area Chair for ICML, NeurIPS, and ICLR. She is a workshop chair for KDD 2026 and an associate editor for ACM Transactions on Knowledge Discovery from Data (TKDD). Her main research interests are graph machine learning and data mining. In particular, she focuses on developing novel computational algorithms for graph models in various fields; her recent research interests include knowledge graph representation learning and graph neural networks.