Aachen 2019 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

T: Fachverband Teilchenphysik

T 42: Experimentelle Methoden I

T 42.4: Talk

Tuesday, March 26, 2019, 16:50–17:05, ST 1

BAT.jl - A new toolkit for Bayesian analysis — •Cornelius Grunwald, Kevin Kröninger, and Salvatore La Cagnina — TU Dortmund, Experimentelle Physik IV, Deutschland

In all but the simplest cases, performing Bayesian inference can be a computationally challenging task. Performant algorithms and tools are needed to apply Bayesian statistics to the complex problems of modern data analysis. The Bayesian Analysis Toolkit (BAT) is a software package that allows the application of Bayesian inference through the use of Markov Chain Monte Carlo techniques. BAT provides a toolbox of algorithms and statistical methods that facilitate solving user-defined problems in a Bayesian approach. In order to improve its applicability, a complete rewrite of BAT is currently in process. Using the Julia programming language for core development and targeting innovative sampling algorithms, the performance of BAT will be increased and computation times will be reduced. Software dependencies will be minimized and interfaces to commonly used languages and tools are going to be provided. This will enhance the general usability of BAT and allow to extend its applicability to further fields of research. In this talk, the approach for the redevelopment of BAT will be presented and first insights into the current status of the project will be given.