David Yallup

Research Associate,

University of Cambridge

I am a research associate based at the Kavli Institute for Cosmology, Department of Physics, University of Cambridge. I am an early career researcher aiming to bring together a mix of interests in ML and precision fundamental science. This page details my current research interests, some biographical information and most importantly contains links to contact me. Please reach out if you are interested in collaborating on any of the topics listed.

My fundamental goal is to find out how we can best build explainable and equitable AI, and I like to consider a number of scientific applications of the technology alongside this.

Education

PhD in Particle Physics, 2019
University College London
MSc in Particles, Strings and Cosmology, 2015
Durham University
MSci in Natural Sciences, 2013
Durham University

Interests

Bayesian Machine Learning
Particle Physics
Cosmology

Research Themes

Neural Anomaly detection for Collider physics

Further details of proposed Cambridge part III project on Anomaly detection

Bayesian Symbolic regression

Further details of proposed Cambridge part III project on Symbolic Regression

Marginalised Gaussian Processes for Cosmology

Further details of proposed Cambridge part III project on Gaussian processes

Sampling graph neural networks

Further details of proposed Cambridge part III project on DeepSets

Bayesian Machine Learning

Investigating the interplay between rigorous statistical techniques used in precision science, and inference performed with modern Machine Learning. Can techniques developed in physics help open the black box of Machine Learning?

Cosmology

Enhancing the analyses of next generation experiments with the power of Machine Learning. How do we build analyses fit for the data rich experiments to unlock their potential?

Particle Physics

Examining the interface of Particle Physics (particularly at High Energy colliders) with Bayesian inference techniques. Can we learn something new about the smallest scale physics with Bayesian techniques?

Featured Publications

Hunting for bumps in the margins

David Yallup, Will Handley

November 2022 Prepared for Journal submission

We demonstrate a fully Bayesian pipeline to approach particle physics modelling problems with.

Split personalities in Bayesian Networks

David Yallup, Will Handley, Mike Hobson, Anthony Lasenby, Pablo Lemos

August 2022 Prepared for conference submission

We demonstrate that even the simplest neural networks have truly multimodal solutions and it is only by combining these candidate models that the true probabilistic description can emerge.

arXiv

Exploring phase space with Nested Sampling

David Yallup, Timo Janßen, Steffen Schumann, Will Handley

May 2022 Published in EPJC

We motivate a connection between Bayesian evidence and particle physics cross sections, and use this to introduce a new path for particle physics calculations

DOI arXiv