Abstract:
In the first part of this talk I will give an tutorial on the foundations of Bayesian machine learning. I’ll cover topics such as the rationale for probabilistic modelling, the choice of priors, approximate inference algorithms and selected topics in Bayesian nonparametrics. In the second part of this talk, I will describe the “Automated Statistician”, a project which aims to automate the exploratory analysis and modelling of data. Our approach starts by defining a large space of related probabilistic models via a grammar over models, and then uses Bayesian marginal likelihood computations to search over this space for one or a few good models of the data. The aim is to find models which have both good predictive performance, and are somewhat interpretable. Our initial work has focused on the learning of unknown nonparametric regression functions, and on learning models of time series data, both using Gaussian processes. Once a good model has been found, the Automated Statistician generates a natural language summary of the analysis, producing a 10-15 page report with plots and tables describing the analysis. I will focus in particular on the modelling of time series, including how we handle change points in Gaussian process models. I will also discuss challenges such as: how to trade off predictive performance and interpretability, how to translate complex statistical concepts into natural language text that is understandable by a numerate non-statistician, and how to integrate model checking.
Bio:
Zoubin Ghahramani’s early childhood was spent in the former Soviet Union and Iran. His family then moved to Spain where he attended the American School of Madrid for 10 years. He studied at the University of Pennsylvania where he was given the Dean’s Scholar Award and obtained a BA degree in Cognitive Science and a BSEng degree in Computer Science and Engineering in 1990. In 1995, he obtained his PhD in Cognitive Neuroscience from the Massachusetts Institute of Technology funded by a Fellowship from the McDonnell-Pew Foundation. His dissertation was entitled “Computation and Psychophysics of Sensorimotor Integration” and his PhD advisor was Michael Jordan. He moved to the University of Toronto in 1995 where he was an ITRC Postdoctoral Fellow in the Artificial Intelligence Lab of the Department of Computer Science, working with Geoffrey Hinton. From 1998 to 2005, He was faculty at the Gatsby Computational Neuroscience Unit, University College London.
He is currently a Professor of Information Engineering, at the University of Cambridge, where he leads the activities in the Machine Learning Group and coordinates Cognitive Systems Engineering. He is also an Associate Research Professor in the School of Computer Science at Carnegie Mellon University, and Adjuct Faculty at the Gatsby Unit, University College London and at POSTECH, South Korea.
His current research interests include Bayesian approaches to machine learning, artificial intelligence, statistics, information retrieval, bioinformatics, and computational motor control. He has recently worked on Gaussian processes, non-parametric Bayesian methods, clustering, approximate inference algorithms, graphical models, Monte Carlo methods, and semi-supervised learning.