group

Andrea works on trustworthy machine learning, with the broader goal of developing ML tools that enable social scientists to analyze Dutch population networks and test their hypotheses within the Macroscope project.

Giray studies privacy risks in machine learning models. His broader goal is to develop methods for releasing models and reliability assessments of model decisions with formal privacy guarantees. His work is supported by PriXAI.

Hanwen develops differential geometry methods to better understand the structure and behavior of learned representations in deep neural networks. His broader goal is to use these insights to improve how we interpret model behavior and develop reliability assessments for model decisions. His work is supported by PriXAI.

Topic: Accelerating Deep RL

Yuan studies more efficient deep reinforcement learning, including graph-based abstractions for improving sample efficiency.

Topic: Characterizing Learning Difficulty in Graph-Structured Data

Tianqi’s PhD examined how data and task complexity shape learning on graph-structured data, with a focus on robustness and reliability.

Topic: Privacy-Preserving Graph Machine Learning

Emmanuel studied privacy risks in graph learning and developed methods to quantify and mitigate them under privacy guarantees.

Topic: Joint Learning from Multiple Information Sources for Biological Problems

Ngan developed methods for learning jointly from omics, clinical, and biological network data to improve generalization in biomedical settings.