This course will introduce the basic concepts of probabilistic graphical models. Graphical Models are a unified framework that allow to express complex probability distributions in a compact and computationally tractable way. Many machine learning applications are tackled by the use of these models, in this course we will highlight the possibilities with applications mainly from natural language processing and computer vision, but potentially also from the life sciences.
The main goal of the class is to understand the concepts behind graphical models and to give hands-on knowledge such that one is able to design models for different applications. The lecture material is roughly divided into three parts: classical graphical models (model classes, factor graph representations, parameter learning, exact and approximate inference techniques), deep generative models (GANs, Variational Autoencoders), and practical applications (image denoising, topic modeling, image generation, ...).
The exercises will be a mix of theoretical and practical assignments. Instead of a final exam there will be a final project (depending on the number of participants to be solved alone or in small teams).

Target group: students interested in using probabilistic graphical models for their research

Prerequisites: * strong quantitative background (linear algebra, calculus, probabilities)
* being able to read code in Python
* being able to program in a language that allows scientific numerical computing, ideally Python.
* having taken the statistical machine learning and/or DSSC core courses are a plus

Evaluation: 50% homework, 50% final project

Teaching format: lectures, homework, final project/online

ECTS: 3 Year: 2020

Track segment(s):
CS-AI Computer Science - Artificial Intelligence
DSSC-PROB Data Science and Scientific Computing - Probabilistic Models

Teacher(s):
Paul Henderson Christoph Lampert

Teaching assistant(s):

If you want to enroll to this course, please click: REGISTER