Accelerated Bayesian inference using deep learning

Add to your list(s) Download to your calendar using vCal

• Adam Moss (Nottingham)
• Wednesday 12 June 2019, 11:00-11:30
• Battcock Tea area.

If you have a question about this talk, please contact Will Handley.

I introduce a novel Bayesian inference tool that uses a neural network to parameterise efficient Markov Chain Monte-Carlo (MCMC) proposals. The target distribution is first transformed into a diagonal, unit variance Gaussian by a series of non-linear, invertible, and non-volume preserving flows. Neural networks are extremely expressive, and can transform complex targets to a simple latent representation from which one can efficiently sample. Using this method, I develop a nested MCMC sampler, finding excellent performance on highly curved and multi-modal analytic likelihoods. I also demonstrate it on Planck 2015 data, showing accurate parameter constraints, and calculate the evidence for simple one-parameter extensions to LCDM in ~20 dimensional parameter space.

This talk is part of the Cavendish Astrophysics Coffee talks series.

This talk is included in these lists:

• All Cavendish Laboratory Seminars
• All Talks (aka the CURE list)
• Battcock Tea area
• Cavendish Astrophysics Seminars
• Centre for Health Leadership and Enterprise
• Combined External Astrophysics Talks DAMTP
• Cosmology, Astrophysics and General Relativity
• Featured lists
• ME Seminar
• Neurons, Fake News, DNA and your iPhone: The Mathematics of Information
• School of Physical Sciences
• Thin Film Magnetic Talks

Note that ex-directory lists are not shown.