Generative Modelling

In this course, we will explore the theoretical and technical underpinnings of a range of generative models. We will start with a comprehensive study of fully-observed models, including Autoregressive models, Random Fields, Markov Models, and more. We will delve into the latent variable models such as probabilistic PCA, variational autoencoders and relevant re-parameterization techniques. Additionally, we will cover energy-based models like restricted Boltzmann machines, deep belief networks, as well as score-based models like diffusion models. Relevant inference methods such as maximum likelihood, Monte Carlo methods, exact methods, Laplace approximation, maximum a posteriori estimation, variational inference, and Markov Chain Monte Carlo, will be introduced. 

Furthermore, we will examine the practical applications that utilize generative modeling, particularly in the domains of Image Synthesis, Natural Language Processing (NLP), and multimodal learning.

Through this course, students will gain an understanding of generative models, contrast their strengths and weaknesses, and evaluate them for real-world tasks in various domains.

The final grade of the course consists of the following components:

• Digital Exam (weighting 60%)

• Group Presentation : proposal project (weight 20%)

• Group Report: final project report (weighting 20%)

Final grade calculation = 0.7 Digital Exam + 0.20 Group presentation + 0.20 * Group Report

A passing final grade for the course can only be earned when for all components at least a 5.0 is earned, and the weighted final grade is at least a 5.8.

In case of an insufficient final result, repair options may exist in accordance with Article 17A : Times and number of examinations, sub 1, of the Teaching and Examination Regulations, for:

• Digital Exam: Resit opportunity

• Group presentation: Repair opportunity

• Group report: Repair opportunity