Gaussian process classification
(30 minutes to learn)
Summary
Gaussian process classification is a Bayesian model for nonparametric classification. The data points have associated latent variables which are drawn from a GP prior, and the labels are modeled as stochastic functions of the latent variables.
Context
This concept has the prerequisites:
- Gaussian process regression (Gaussian process classification builds on Gaussian process regression.)
- Bayesian logistic regression (Gaussian process classification is a kernelized version of Bayesian logistic regression.)
Core resources (read/watch one of the following)
-Free-
→ Gaussian Processes for Machine Learning
A graduate-level machine learning textbook focusing on Gaussian processes.
Location:
Section 3.3, pages 39-41
-Paid-
→ Pattern Recognition and Machine Learning
A textbook for a graduate machine learning course, with a focus on Bayesian methods.
Location:
Section 6.4.5, pages 313-315
See also
- We can approximate the solution to GP classification using the Laplace approximation .
- Gaussian processes have a variety of uses in machine learning, including:
- regression
- black-box optimization (where we only get to evaluate the function, and doing so is expensive)
- reinforcement learning