PyAutoFit: A Classy Probabilistic Programming Language For Data Science

Classy Probabilistic Programming For Big Data

James Nightingale

Big Data Data Science Open-Source Science

See in schedule: Fri, Jul 30, 13:45-14:15 CEST (30 min) Download/View Slides

A major trend in data science is the rapid adoption of Bayesian statistics for data analysis and modeling. With modern data-sets growing by orders of magnitude in size, the focus is now on developing methods capable of applying contemporary inference techniques to extremely large datasets. To this aim, I present PyAutoFit (https://github.com/rhayes777/PyAutoFit), an open-source probabilistic programming language for automated Bayesian inference that was recently published in the Journal of Open Source Software (https://joss.theoj.org/papers/10.21105/joss.02550).

I will begin by giving an overview of PyAutoFit’s core features, in particular how it:

- Makes it simple to compose and fit probabilistic models using a range of Bayesian inference libraries, such as emcee (https://github.com/dfm/emcee) and dynesty (https://github.com/joshspeagle/dynesty).

- Handles the 'heavy lifting' that comes with model-fitting, including model composition & customization, outputting results, model-specific visualization and posterior analysis.

- Is built for big-data analysis, whereby results are output as a sqlite database which can be queried after model-fitting is complete.

PyAutoFit was developed by Astronomers seeking to fit large libraries of galaxy images to better understand the nature of dark matter. Using this science-case, I will describe PyAutoFit’s advanced features, such as multi-level models, automated model-fitting pipelines and support for massively parallel computing infrastructures.

The goal of this talk is to introduce the audience to PyAutoFit so they can adopt it for their use-case. The only prerequisite is a basic understanding of object oriented programming in Python.

Type: Talk (30 mins); Python level: Beginner; Domain level: Beginner