Graph- and Physics-Informed Learning for Multi-Domain Scientific Discovery

Speaker: Dr. Qianru ZHANG
Post-doctoral Fellow
Yale University

Title: Graph- and Physics-Informed Learning for Multi-Domain Scientific Discovery

Date: Tuesday, 19 May 2026

Time: 11:00am - 12:00noon

Venue: Room 1409 (via lift 25/26), HKUST

Abstract:

Data across biological, urban, and social domains are inherently complex, dynamic, and non-Euclidean. A fundamental question in modern AI research is: How can we design AI models that effectively and efficiently represent these diverse, interconnected domains? In the talk, Dr. Zhang will present a unified AI framework that transcends traditional domain silos through graph- and physics-informed learning. First, in the realm of biological and physical sciences, Dr. Zhang will introduce quantum-informed AI models (such as OG-QIMP) that learn directly from fundamental physical and chemical principles rather than mere empirical correlations. This approach enables stronger generalization and robustness across unseen molecular distributions. Second, in urban science, Dr. Zhang will discuss certified, efficient, and generalizable AI frameworks for spatial-temporal forecasting. These models address the paradigm shift from empirical approximation to formally verified reasoning, ensuring mathematically verified model reliability. Finally, in social science, Dr Zhang will demonstrate how hyperbolic state-space modeling (Hyperbolic Mamba) effectively captures hierarchical social structures and long-term behavioral sequences. Together, these works establish theoretical and algorithmic bridges that unify physics-informed and graph-based representations, pioneering a vision toward universal, cross-domain scientific intelligence grounded in structure and physical laws.

Biography:

Dr. Qianru ZHANG is currently a Post-doctoral Fellow at Yale University. She received her Ph.D. in Computer Science from The University of Hong Kong in 2024. She also held postdoctoral position at the University of Cambridge in 2025. Dr Zhang’s research focuses on developing physics-informed, knowledge-enhanced AI that bridges symbolic reasoning and scientific discovery across biological, physical, urban, and social sciences. She has a prolific publication record, with over 24 peer-reviewed papers in top-tier conferences and journals, including ICML, NeurIPS, ICDE, WWW, TKDE, and TOIS. Her pioneering work has led to multiple invitations for keynote and guest talks at prestigious institutions such as Cambridge, Yale, and Cornell.