About Me
-
Curriculum Vitae (updated May 2022)
[pdf] -
My Name
My given name is "Yee Whye" (pronounced as an E followed by a Y), and family name is "Teh". It is a Malaysian Chinese (Teochew) name. The Chinese characters are "郑" for "Teh" and "宇怀" for "Yee Whye". -
Academic Bio
I am a Professor of Statistical Machine Learning at the Department of Statistics, University of Oxford, a Principal Research Scientist at DeepMind, an Alan Turing Institute Faculty Fellow and an ELLIS Fellow, co-director of the ELLIS Robust Machine Learning Programme and co-director of the ELLIS@Oxford ELLIS unit. I obtained my PhD at the University of Toronto, and did postdoctoral work at the University of California at Berkeley and National University of Singapore, where I was a Lee Kuan Yew Postdoctoral Fellow. I was a Lecturer and a Reader at the Gatsby Computational Neuroscience Unit, UCL and an ERC Consolidator Fellow.
My research interests are in machine learning, computational statistics and artificial intelligence, in particular probabilistic models, Bayesian nonparametrics, large scale learning and deep learning. I also have interests in using statistical and machine learning tools to solve problems in genetics, genomics, linguistics, neuroscience and artificial intelligence.
I was programme co-chair of the International Conference on Artificial Intelligence and Statistics 2010, Machine Learning Summer School 2014 (Iceland), and the International Conference on Machine Learning 2017, an editor for a IEEE TPAMI Special Issue on Bayesian nonparametrics, and is/was an associate/action editor for Bayesian Analysis, IEEE Transactions on Pattern Analysis and Machine Intelligence, Machine Learning Journal, Journal of the Royal Statistical Society Series B, Statistical Sciences and Journal of Machine Learning Research. I have been area chair or senior area chair for NIPS, ICML and AISTATS on multiple occasions. -
Research Interests
My research interests lie in the general areas of machine learning, Bayesian statistics and computational statistics. Although my group works on a variety of topics ranging from theoretical, through to methodological and applications, I am personally particularly interested in three (overlapping) themes: Bayesian nonparametrics and probabilistic learning, large scale machine learning, and deep learning.
These themes are motivated by the phenomenal growth in the quantity, diversity and heterogeneity of data now available. The analysis of such data is crucial to opening doors to new scientific frontiers and future economic growth. In the longer term, the development of general methods that can deal with such data are important testing grounds for general-purpose intelligent computational systems and general artificial intelligence.
Firstly, I am interested in developing Bayesian nonparametric models and methodologies. These are highly flexible models that can accommodate the diversity and heterogeneity of data, and can sidestep model selection, one of the core difficulties of parametric machine learning approaches. They are probabilistic models built out of infinite-dimensional stochastic processes, and have a range of elegant and useful statistical properties, for example power-law behaviours, partial exchangeability and large coverage. I am interested in developing novel Bayesian nonparametric models, inspired by applications to unsupervised learning, computational linguistics and computational biology, and novel algorithms for efficient learning of such models.
Secondly, I am broadly interested in probabilistic approaches to learning and other flexible learning frameworks like deep learning and kernel methods. I think that flexible modelling is key in allowing data to speak for themselves.
Thirdly, I am interested in developing novel algorithms and architectures for large scale machine learning, adapted to modern computational environments which include a wide range of resources with different properties, e.g. GPUs, parallel architectures, and distributed and cloud computing. I am particularly interested in pushing the frontiers of Monte Carlo based learning techniques.
I am known for my work in Bayesian nonparametrics, deep learning and inference and learning methodologies using variational and Monte Carlo techniques.