## PyMC Web Series: Introduction to Hilbert Space GPs in PyMC: A fast Gaussian process approximation that you can actually use

**Speaker:** Bill Engels (@bwengals)

**Event type:** Live webinar

**Date:** March 15 2023 (subscribe here for email updates)

**Time:** 22:00 UTC

**Register for the live webinar** on Meetup to get the Zoom link

## Sponsor

We thank our sponsors for supporting PyMC and the PyMCon Web Series. If you would like to sponsor us, contact us for more information.

Mistplay is the #1 Loyalty Program for mobile gamers - with over 20 million users worldwide. Millions of gamers use our platform to discover games, connect with friends, and earn awesome rewards. We are a fast growing profitable company, recently ranked as the 3rd fastest growing technology company in Canada. Our passion to innovation drives our growth across the industry with the development of new apps, powerful ad tech tools, and the recent launch of a publishing division for mobile games.

Mistplay is hiring for a Senior Data Scientist (Remote or Montreal,QC).

### Content

Welcome to another event in the PyMCon Web Series! As part of this series, most events will have one or more asynchronous components and a at least one live, synchronous component.

#### Recording of Bill’s synchronous talk

#### Bill’s Interview

In this case, Bill has provided:

- A set resources for those looking for a background on Gaussian processes: see here
- A notebook including some pre-made approaches to data analysis for you to try out on your own data. These will include both Gaussian processes, but also related approaches (regression, B-splines, etc.): see here

In addition, approximately a week after Bill’s synchronous talk, we will host a special edition of PyMC Office Hours specifically focused on Gaussian processes of all flavors: see here

**Abstract of the talk**

Gaussian processes (GPs) are a versatile tool in the Bayesian modelers toolbox – in theory. In practice, for all but the smallest data sets, one needs to resort to approximations to actually fit GPs in any reasonable amount of time. There are currently a few GP approximations implemented in PyMC based on inducing points that are fast, but only apply when the likelihood is Gaussian. The Hilbert Space Gaussian Process (HSGP) approximation works well with any likelihood and scales as O(nm + m). In this talk I’ll introduce a PyMC HSGP implementation and show via case studies how it fills a few key gaps in the PyMC GP library: fast GPs as model subcomponents, and fast GPs with non-Gaussian likelihoods. I’ll also cover tips and tricks for applying HSGPs effectively in practice.