This tutorial is on Gaussian Processes for Regression: Models, Algorithms, and Inference. It took place at the 5th Computational Physics School for Fusion Research (CPS-FR 2024) and was hosted by the MIT Plasma Science and Fusion Center. It ran on August 21 and 22. See this link for the latest versions of all tutorials.

Part I: Wednesday August 21, 1:30pm–2:30pm
Part II: Wednesday August 21, 2:45pm–3:45pm
Part III: Thursday August 22, 10:15am–11:15am
Part IV: Thursday August 22, 11:30am–12:30pm

The instructor is Prof. Tamara Broderick. Contact information can be found here.

Prerequisites: It will be helpful to have basic familiarity with (1) Bayesian data analysis and its goals and (2) both univariate and multivariate Gaussian distributions though we will review key facts.

Slides:

• Part I: Gaussian process model [slides]
• Part II: Gaussian process regression [slides]
• Part III: Squared exponential kernel and observation noise [slides]
• Part IV: What uncertainty are we quantifying? [slides]
• A list of resources: [slide]

Code for demos:

• Code for demos from demos from all parts (in order) can be found [here]

Errata:

• Slides 8 erroneously uses superscripts instead of subscripts. I'll fix that in the next few days and post here.
• There are some parts in the code that need cleaning up, but to the best of my current knowledge, there are not actual errors. Please let me know if you find any though!

Accessibility
Plain Academic