Tutorial: Fundamentals of Bayesian Inference using Probabilistic Programming

Welcome to the world of probabilistic programming and Bayesian inference!

In this short course/tutorial (2 x 3h), you will learn the fundamentals of Bayesian inference using a relatively new paradigm called probabilistic programming. No prior knowledge of Bayesian theory is necessary. You will learn about the basic intuitions behind popular inference algorithms and how to design and write small models / probabilistic programs in popular probabilistic programming languages.

Lecturer: David Broman, KTH Royal Institute of Technology
Time: June 1, 14.00-17.00 (CEST) and June 7, 14.00-17.00 (CEST), 2022.
Place: Digital Futures hub, KTH Campus, Stockholm, and over Zoom.
Registration: See this link for course info and course registration.  

 

---

Day I - Basics of Bayesian Statistics and Simple Modeling in WebPPL

Agenda

• 14.00 - 14.10 Welcome and mingling
  
• 14.10 - 15.20 Lecture with mini exercises
  
• 15.20 - 15.40 Coffee break
  
• 15.40 - 17.00 Lecture with mini exercises

[Slides Day 1]

Main topics of day 1

• Introduction to Probabilistic Programming Languages
• Basic Probability Theory
• Bayesian Statistics
• Sampling
• Modeling in WebPPL

---

Day II - Bayesian Inference Methods and Modeling in Stan

Agenda

• 14.00 - 14.10 Welcome and mingling
  
• 14.10 - 15.20 Lecture with mini exercises
  
• 15.20 - 15.40 Coffee break
  
• 15.40 - 17.00 Lecture with mini exercises

[Slides Day 2]

Main topics of day 2

• Rejection and Importance Sampling
• Sequential Monte Carlo
• Markov chain Monte Carlo
• Conjugate priors and Delayed Sampling
• Modeling and inference in Stan
• Latent Dirichlet Allocation (LDA)

Installation of Stan
Before the tutorial starts, it is good if you have the possibility to install Stan on your computer. Please install using the instructions on the Stan website, or do the following simple steps to install PyStan:

• Install Python 3 on your system.
• Install PyStan and packages for visualization:

  python3 -m pip install pystan pandas seaborn

• Download file hello-coin.py and run

  python3 hello-coin.py

• If everything works, a plot file

  plot.pdf

  should be available in the same folder where you ran the Python script.

---

This is a personal web page. More information.