Greg Ver Steeg

Greg Ver Steeg is an Associate Professor in the Computer Science & Engineering Department at University of California, Riverside (UCR), with an Adjunct Research Associate Professor title at University of Southern California (USC). He is also a Part-Time Visiting Academic at Amazon Alexa AI (2023). His research bridges machine learning, information theory, and complex systems, focusing on fair representation learning, diffusion models, and causality.

• Education: Ph.D. in Physics from Caltech

Research Interests: His work spans:

• Information-Theoretic Fairness: Removing group-identifying information to ensure algorithmic fairness with guarantees on accuracy trade-offs.
• Diffusion Models: Connecting log likelihood ratios and classification for robust generative modeling.
• Graph Neural Networks: Innovations like MixHop for neighborhood weighting in graphs.
• Neuroimaging: Federated learning for MRI harmonization across sites.
• Complex Systems: Macro-causal discovery in social and biological data through coarse-graining.

Scientific Contributions: Key papers include the Information Sieve (ICML 2016) for hierarchical factor discovery, MixHop (ICML 2019) for graph convolutional networks, and information-theoretic generalization bounds (ICML 2020) for label noise correction. His lab has produced multiple best paper awards and high-impact preprints on diffusion bridges and material science.

Advising: Mentored students like Rob Brekelmans (ICML Best Paper), Sami Abu-el-Haija (Google Research), and Hrayr Harutyunyan (NeurIPS 2021), with ongoing supervision of Xianghao Kong (ICLR 2023/2024) and Shaorong Zhang (diffusion bridges).