“Learning and Inference in Graphical Models”
Dr. M. Lauer
Graphical models are a technique in machine learning and probability theory to describe complex relationships between different random variables in a comprehensive way. For these models powerful learning and inference techniques have been developed during the last two decades. Application areas for graphical models can be found in modern machine learning, computer vision, and sensory data analysis, among others.
The lecture introduces the foundations of graphical models and various techniques for inference in graphical models. This includes belief propagation, variational Bayesian inference, Markov chain Monte Carlo approaches, the EM/ECM-algorithm, and techniques for inference in models of variable size. The lecture starts with a short repetition of probability theory and Bayesian statistics.