The class provides an introduction to Bayesian nonparametric statistics and machine learning. Class notes are available (see below).

## Class specs

**Time**: Wed 10:00-12:00
**Room**: 1025 SSW

## Abstract

A Bayesian model is called nonparametric if its parameter space is a functional space (e.g. a space of measures or continuous functions). Since a Bayesian model defines a probability distribution on its parameter space, and a distribution on a functional space is a stochastic process, Bayesian nonparametrics is, very roughly speaking, Bayesian inference for stochastic processes. Statistical applications include problems such as clustering, matrix factorization, and Gaussian process regression. We will cover topics including:

- The fundamental models, in particular the Dirichlet process and the Gaussian process.
- Some technical aspects of Bayesian models with infinite-dimensional parameter spaces.
- Algorithmic inference.
- Exchangeable random structures and their application to Bayesian modeling problems.

## Material

- Lecture notes
- Slides on sampling algorithms
- Bayesian nonparametrics tutorial page with references