Full Professor, University of Texas, Dallas
2 papers at NeurIPS 2025
TopER is a low-dimensional, interpretable graph embedding method based on topological evolution rates, enabling intuitive visualization and strong performance on clustering and classification tasks.
We introduce the first benchmark of 84 real-world temporal graphs and MiNT, the first multi-network temporal model pre-trained on this collection, demonstrating strong transferability to unseen token networks.
