This is the twentieth lecture in the Probabilistic ML class of Prof. Dr. Philipp Hennig, updated for the Summer Term 2021 at the University of Tübingen. Slides available at https://uni-tuebingen.de/en/180804. Contents: * How to design probabilistic machine learning solutions * Latent Dirichlet Allocation * conditional independence (rejoinder) * Gibbs sampling (rejoinder) © Philipp Hennig / University of Tübingen, 2021 CC BY-NC-SA 3.0