Bayesian Data Science Two Ways: Simulation and Probabilistic Programming | SciPy 2018 Tutorial

In this tutorial, Hugo and Eric will build participants’ knowledge of Bayesian inference, workflows and decision making under uncertainty. We will start with the basics of probability via simulation and analysis of real-world datasets, building up to an understanding of Bayes’ theorem. We will then introduce the use of probabilistic programming to do statistical modelling. Throughout this tutorial, we will use a mixture of instructional time and hands-on time. During instructional time, we will use a variety of datasets to anchor our instruction; during hands-on time, which immediately follow instructional time, our participants will apply the concepts learned to the Darwin’s finches dataset, which will permeate the entire tutorial.
For maximal benefit, participants are expected to have experience writing for-loops and working with numpy arrays. The necessary syntax will be introduced during the tutorial, but familiarity will help. No knowledge of statistics is necessary. However, the most important and beneficial prerequisite is a will to learn new things. If you have this quality, you'll definitely get something out of this tutorial!

[arunsupe]

The content is excellent. The two presenters have done a very good job. They communicate the intuition using simple simulations initially. This is helpful for understanding. They then introduce pymc3 as the tool that abstracts away the details and provides an API for building generative models.



The presenters are knowledgeable; there is no inconsistency in their statements. Makes this a high quality presentation. And funny, especially Hugo... his reason for setting the random seed to 42.


Thanks to Hugo, Eric, Enthought & scipy for this presentation.

[perrygrossman]

"at home or wherever else you like to hang That was weird..." FTW! At 1:17:00.