- Machine Learning for Graphs (WS 2021-22 and SS 2020 at Leibniz University Hannover): Machine learning for graphs (MLG), whereby classical ML is generalized to irregular graph domains, is emerging as a fundamental methodology to process graph structured data in various disciplines. The course was designed along with my colleague Thorben Funke to cover the following learning objectives: (i) recognize the challenges in learning from graph structured data (ii) explain, compare and implement various shallow and deep learning techniques for graph data (iii) explain interpretability techniques for graph ML.
Previously in Summer 2020 I also offered a seminar on Machine Learning for Graphs. In this seminar we discuss recent papers in machine learning algorithms for graph data. The students gain skills in understanding, presenting and reporting advanced scientific works.
- Probabilistic Machine Learning (WS 2019-20 and WS 2020-21 at Leibniz University Hannover): Machine Learning can be defined as a set of methods to automatically detect patterns in data which can be used to make predictions in future data. A key concept in this field is uncertainty which arises both through noise measurements and through the finite size of the datasets. The main learning objectives of this course were, such that (i) Students understand and are able to apply the mathematical abstractions to express uncertainty in their model design. (ii) Students learn how to analyze the models’ results, improve the model parameters and can interpret the model predictions and their relevance.
The course contents were based on the book Pattern Recognition and Machine Learning by Christopher M. Bishop and include Bayesian approaches to Linear Regression and Classification, neural networks, graphical models, expectation maximization and clustering, methods for approximate inference.