Teaching

The term optimization refers to techniques for the identification of the best solution in a complex problem setting. Many applications from machine learning and engineering are based on solving an optimization problem. This lecture introduces the major theoretical and algorithmic tools for solving of convex optimization problems. Practical problems for machine learning, engineering and further application domains are used as illustration. The students apply their knowledge to practical optimization problems in tutorial exercises.

This lecture covers key concepts in pattern recognition and classification, focusing on feature types, transformations, and dimensionality reduction methods such as PCA and ICA. It introduces Bayesian decision theory and methods for estimating class probabilities and parameters. Both parametric and non-parametric classification techniques are presented, including maximum likelihood, k-NN, and Parzen windows. A range of classifiers is discussed, such as SVMs, decision trees, perceptrons, and classifiers for sequences and nominal features. The lecture also addresses challenges like overfitting and explores learning principles, performance evaluation, and boosting.

This lecture covers the acquisition, description, processing, and evaluation of image data for the purpose of automatic visual inspection. Various sensors and methods for capturing image-based data, as well as the relevant optical principles, are discussed. The mathematical description of image signals is examined in detail. The necessary system-theoretical methods and relationships are derived and discussed. The second half of the lecture focuses on the various sub-tasks and all major signal processing methods used in digital image processing and analysis.

This is a Seminar offered for master’s students. The Seminar is offered every summer semester, with me supervising one topic in every session.

This is a Proseminar offered for bachelor’s students. The Proseminar is offered each semester, with me supervising one topic in every session.

I supervised a total of 10 Master’s theses during my PhD:

• Bin-Based Learnable Sampling Strategy for Point Cloud Edge Sampling, Hao fu, 2024.
• A Study of Attention Module Variations in Point Cloud Feature Learning, Kaige Wang, 2023.
• Exploring Contrastive Learning on 3D Point Clouds with Cross Attention, Qianlinag Huang, 2023.
• Attention-based 3D Point Cloud Segmentation on Bosch Motors and Its Visualization Interface Design, Xuelei Bi, 2022.
• VoxAttention: Shape Synthesis via Part Assembly with Self-Attention, Junwei Zheng, 2022.
• A Novel Multi-Attribute Regression Benchmark Suite on Bosch Motors, Linxi Qiu, 2022.
• 3D Point Cloud semantic Segmentation on BOSCH Motors, Haodong Yu, 2021.
• SwitchVAE: Learning Better Latent Representations from Objects of Different 3D Euclidean Formats, Mingyuan Zhou, 2021.
• Improving Stylized View Synthesis of Image-based Reconstructions using Neural Networks, Iris Mehrbrodt, 2019.
• Object Detection In 3D Point Clouds By Leveraging Local-Spatial Correlations, Kangning Li, 2019.