Peter Orbanz

Peter Orbanz is a Professor of Machine Learning at the Gatsby Computational Neuroscience Unit, University College London (UCL). He previously held roles as Assistant and Associate Professor in the Department of Statistics at Columbia University, and completed postdoctoral research at the University of Cambridge under Zoubin Ghahramani. His research focuses on machine learning, Bayesian nonparametrics, symmetry in statistical models, and dynamics. He has spent sabbaticals at the Isaac Newton Institute, Microsoft Research New England, and UC Berkeley.

Orbanz earned his PhD in 2008 from ETH Zurich under Joachim M. Buhmann, with a thesis on Bayesian nonparametric models. His work bridges theoretical foundations of machine learning with practical applications, emphasizing symmetry properties and scalable inference methods. Current research includes causal uncertainty quantification, Gaussian process dynamics, and equivariant neural networks.

He teaches advanced courses on probabilistic learning, Bayesian nonparametrics, and statistical theory at UCL and Columbia. His research group at UCL includes PhD students Hugh Dance, Kevin H. Huang, Vasco Portilheiro, and Vince Velkey. Former advisees include Morgane Austern (Harvard), Benjamin Bloem-Reddy (UBC), and Victor Veitch (Chicago).

Orbanz has authored influential papers on topics such as exchangeable random structures, Gaussian universality, and scalable MCMC methods. His work often integrates concepts from probability theory, statistical physics, and computational neuroscience. He maintains active collaborations with institutions worldwide and contributes to open-source tools for Bayesian modeling.