I’m interested in the theory and practice of understanding and building systems that learn and make decisions. Humans have an exceptional ability to learn from experience, which sets them apart from current artificial intelligent (AI) systems. To understand human learning and design better AI we need principled approaches to learning and decision making based on Bayesian inference in machine learning. My interests span: probabilistic inference, reinforcement learning, approximate inference (variational and MCMC), decision making, non-parametric modeling, stochastic processes and efficient learning.
Gaussian processes (GPs) are a principled, practical, probabilistic approach to learning in flexible non-parametric models. GPs have found numerous applications in regression, classification, unsupervised learning and reinforcement learning. Great advances have been made recently in sparse approximations and approximate inference. My book Gaussian Processes for Machine Learning, MIT Press 2006, with Chris Williams is freely available online. I also maintain the gpml matlab/octave toolbox with Hannes Nickisch, as well as the pretty outdated Gaussian Process website.