webinar register page

Webinar banner
Chris Oates and Takuo Matsubara - A Covariance Function Approach to Prior Specification for Bayesian Neural Networks
Bayesian neural networks attempt to combine the strong predictive performance of neural networks with formal quantification of uncertainty associated with the predicted output in the Bayesian framework. However, it remains unclear how to endow the parameters of the network with a prior distribution that is meaningful when lifted into the output space of the network. A possible solution is proposed that enables the user to posit an appropriate covariance function for the task at hand. Our approach constructs a prior distribution for the parameters of the network that approximates the posited covariance structure in the output space of the network. We discuss the pros and cons of this approach and highlight some interesting directions for future work.

Oct 21, 2020 05:00 PM in Paris

Webinar logo
* Required information
Loading

Speakers

Chris Oates
Professor of Statistics @Newcastle University
Chris Oates is a Professor of Statistics at Newcastle University and a Group Leader for the Programme on Data-Centric Engineering at the Alan Turing Institute, UK. His research interests include computational statistics, inverse problems, kernel methods, probabilistic numerical methods and uncertainty quantification. He obtained his PhD from the University of Warwick in 2013, where he continued as a research fellow before moving to the University of Technology Sydney in 2015 and finally Newcastle University in 2017.
Takuo Matsubara
PhD student @School of Mathematics, Statistics and Physics, Newcastle University
Takuo Matsubara is a PhD student in the School of Mathematics, Statistics and Physics, Newcastle University and the Alan Turing Institute, UK. His research interests include Bayesian statistics, statistical machine learning, and theory and application of reproducing kernel Hilbert spaces. He obtained M.Eng. and B.Eng. degrees in the Department of Electrical Engineering and Bioscience, Waseda University in Tokyo, Japan, and prior to his PhD he worked as a research assistant at the RIKEN Center for Advanced Intelligence Project