๐๏ธ Week 08 - Gaussian process
This week, we will explore Gaussian processes (GPs), a powerful and flexible framework for modeling complex data. We will develop GPs from two perspectives: as a prior distribution over functions and as a Bayesian regression model on implicitly defined basis functions. We will discuss the properties of GPs, including their mean and covariance functions, and how they can be used for regression. We will also cover the concept of hyperparameters and model selection and how to optimize them using maximum likelihood estimation. Finally, the last part of the lecture will be dedicated to a overview of challenges when working with GPs, including computational issues and non-Gaussian likelihoods (e.g. for classification problems).
๐ Lecture Slides
Print-friendly version
๐ Recommended Reading
๐ References
Murphy, Kevin P. 2023. Probabilistic Machine Learning: Advanced Topics. MIT Press. http://probml.github.io/book2.
Rasmussen, Carl Edward, and Christopher K. I. Williams. 2005. Gaussian Processes for Machine Learning (Adaptive Computation and Machine Learning). The MIT Press.