David Widmann

I am a PhD student at the Division of Systems and Control within the Department of Information Technology and the Centre for Interdisciplinary Mathematics in Uppsala, supervised by Fredrik Lindsten, Dave Zachariah, and Erik Sjöblom. The main focus of my PhD studies is uncertainty-aware deep learning. Currently, I am particularly interested in analyzing and evaluating calibration of probabilistic models. Please visit my webpage for more information.

My GitHub profile provides an overview of my contributions to the Julia ecosystem. Currently, I am a member of the steering council of SciML and the Turing team.

The speaker's profile picture


EllipticalSliceSampling.jl: MCMC with Gaussian priors
David Widmann

EllipticalSliceSampling.jl is a package for elliptical slice sampling (ESS), a simple Markov chain Monte Carlo method for models with Gaussian priors. Models can be specified with an arbitrary Julia function of the log-likelihood and a Gaussian prior that follows a simple interface (Distributions.jl works out of the box). Features such as progress logging and parallel sampling are provided automatically by AbstractMCMC.jl. Turing.jl supports ESS, also within Gibbs, through this package.

Calibration analysis of probabilistic models in Julia
David Widmann

Calibrated probabilistic models ensure that predictions are consistent with empirically observed outcomes, and hence such models provide reliable uncertainty estimates for decision-making. This is particularly important in safety-critical applications. We present Julia packages for analyzing calibration of general probabilistic predictive models, beyond commonly studied classification models. Additionally, our framework allows to perform statistical hypothesis testing of calibration.