An Introduction to Variational Bayes - online

Variational Bayes is a tool used to allow scalable Bayesian inference and is applicable in a huge variety of applications. This course will give an introduction to Variational Bayes and describe its use in some common regression settings. The course will consist of a mixture of taught lectures and R practicals, giving participants an understanding of both some of the theory of variational methods, and the practical implementation in some common models.

Course Description

 

2-day course structure as follows:

Day 1 (10th October):  

• 10-11: Lecture 1 – An overview of Variational Bayes 
• 11.15-12.15: Lecture 2 – Variational Bayes for common regression models 
• 12.30-1.00: Group practical activity and questions
• Lunch
• 2.00-3.00: Lecture 3 – Issues and extensions with Variational Bayes 
• 3.15-3.45: Group practical activity and questions
• Homework

Day 2 (17th October):  

• 10-12: Practical Q+A – discussion of practical worksheets, and opportunity to discuss matters arising during the course  

 

The taught content will all come on day 1. We will give an overview of variational Bayes methods and discuss some common uses. There will be 2 short practical sessions on day 1 to get going with applying variational Bayes methods. Following this, there will be some set tasks based on more advanced variational Bayes models that participants can work through in their own time. On the second day, there will be a 2 hour Practical clinic which will provide opportunity for candidates to ask questions about the theoretical concepts, and the practical worksheets. Full solutions will be provided after the course.

 

Participants should ensure that they have either R or RStudio installed on their computer. Some prior knowledge of using R will be assumed.