Full Name
Oliver Johnson
Job Title
Professor of Information Theory, School of Mathematics
Company
University of Bristol
Speaker Bio
I am Professor of Information Theory and Director of the Institute for Statistical Science in the School of Mathematics at the University of Bristol. I am Personal Advisor for the COMPASS CDT in Computational Statistics and Data Science and help run the Centre for Doctoral Training in Communications. My book Numbercrunch is written for a general audience and I can be found on Substack.

Much of my research applies Information Theory to understand limit theorems in probability and statistics. Recently this has involved discrete random variables - including transportation of measure, proofs of the Shepp-Olkin entropy concavity and monotonicity conjectures, log-Sobolev inequalities, maximum entropy, monotonicity and other problems.

I also research group testing, both in terms of practical algorithms and fundamental bounds via converse results. This has led to a recent survey monograph, published by Foundations and Trends in Communications and Information Theory.

I published a book about my older research, entitled Information Theory and the Central Limit Theorem. Here is a list of known errata. My preprints and academic links are available on this site.

In the academic year 2023-24, I will be lecturing the first half of the first year unit Probability and Statistics.

I was previously Max Newman Research Fellow of the Statistical Laboratory of Cambridge University and Clayton Fellow of Christ's College Cambridge, Associate Director of the Heilbronn Institute for Mathematical Research, Director of Equality, Diversity and Inclusion for Bristol Mathematics and Programme Director for the MSc in the Mathematics of Cybersecurity. https://people.maths.bris.ac.uk/~maotj/
Abstract
Andrew Barron’s 1986 Annals of Probability paper opened up the study of the information-theoretic Central Limit Theorem, which has been an area of active research ever since. I will describe some of this history, with a particular focus on the role played by projections and Hermite polynomials in establishing properties of the Fisher information. In our 2001 paper, Andrew and I used these ideas to prove an explicit rate of convergence under a spectral gap condition, and I will give an alternative perspective on this in terms of the HGR maximal correlation.
Oliver Johnson