# New PyMCon Talk: The Only Constant is Change - Bespoke Changepoint Modelling in PyMC by Abuzar Mahmood

Welcome to the 11th event of the PyMCon Web Series!

Speaker: Abuzar Mahmood, Neuroscientist, Brandeis University
Event type: Live Talk with Q&A
Live Session Date/Time: 2023-11-20T14:00:00Z (subscribe here for email updates)
Website: PyMCon Events · PyMCon Web Series
Register for Live session: Meetup event (to get the Zoom link)

## Abstract of the Talk:

Dynamic data are all around us. Changepoint models allow us to know when changes happen in these data and what they look like. Probabilistic modelling allows us to elegantly build customizable changepoint models for different data types, as well as provide us with uncertainty estimates for the position and magnitude of the change (both indispensable quantities for decision-making and hypothesis testing). This tutorial will briefly cover building changepoint models for multivariate data using PyMC but will primarily focus on the ways in which this “basic” model can be extended.

This tutorial is targeted towards academic researchers, data scientists, and anyone interested in being able to easily build bespoke models which provide uncertainty estimates for inferred statistics. This talk will attempt to be accessible to beginners but leans towards more intermediate users interested in changepoint modelling. Previous experience with PyMC, and a background in statistical modelling is assumed. No libraries other than PyMC and the basic scientific stack (numpy, scipy, matplotlib) will be used.

The tutorial aims to be hands-on, will discuss some theory to provide context for the models discussed, and will be heavy on understanding code to construct the “guts” of the models (in particular, selection of distributions for modelling the emissions and changepoint locations, and the details of the tensor manipulation to put everything together).

The tutorial will be divided into three parts:

1. Introduction to changepoint modelling and use cases.
2. Construction of the “basic” multivariate changepoint model.
3. Extensions to the “basic” model:
1. Handling drift/noise in repeated observations using mixture emissions across multiple timeseries observations.
2. Determining the distribution over number of changepoints using a Dirichlet Process prior.

## Content

Abuzar’s Interview:

Adia Lab is an independent, Abu Dhabi-based laboratory dedicated to basic and applied research in data and computational sciences.
ADIA Lab focuses on societally-important topics such as climate change and energy transition, blockchain technology, financial inclusion and investing, decision making, automation, cybersecurity, health sciences, education, telecommunications, and space, by conducting cutting-edge research in Data Science, Artificial Intelligence, Machine Learning, and High-Performance Computing.

1 Like

Great news for everyone! Abuzar has pre-recorded in-depth model walkthroughs for you. Now, you can dive into the notebook at your own pace before the live event. It’s the perfect way to gear up and make the most out of the live session!

What’s New?

• Walkthroughs of the models for a head-start.
• Focus on hands-on learning with the provided notebook:

Live Session Highlights:

• Abuzar will discuss the background and context of changepoint modelling live.
• A deep dive into the mixture model and its practical applications.

Don’t miss this opportunity to enhance your learning experience! Prep with the pre-recorded sessions and bring your questions and insights to the live discussion.

Set your reminders and get ready for the live session filled with expert insights and interactive discussions!

Register for Live session (to get the Zoom link)

1 Like

Only a few hours left until our Changepoint Modeling session with Abuzar Mahmood!

Join us at 2023-11-20T14:00:00Z (RSVP for Zoom link) : [PyMCon] The Only Constant is Change: Bespoke Changepoint Modelling in PyMC, Mon, Nov 20, 2023, 9:00 AM | Meetup

Don’t miss out!

I’ve got a question. Is it possible to go back to a previous state?

Screenshot from zoom

What’s the reasoning behind choosing a threshold for numbers of states based on the duration of the dataset? Why choose 10% as a threshold? Is it just data dependent?

And maybe if we look at the ratio of state “lifetimes” (ranked by lifetimes), a similar cutoff would appear just because most of the states have a much smaller lifetime than the others.

1 Like

Great Talk (:

1 Like

In case you missed the live session, the video has been posted to YouTube.

Watch the recording here: