For more information about the seminar, also see the Learnweb page at https://www.uni-hildesheim.de/learnweb2023/course/view.php?id=1062

As most metric learning techniques we will study in the seminar are based on deep neural networks, prior knowledge about deep learning is recommended. Students in Msc Data Analytics should already have passed Data Analytics I.

In the seminar, we will study and discuss so-called metric learning approaches. Metric learning is a subfield of supervised machine learning concerned with building models that embed instances into a vector space, such that instances from the same class are close and instances from different classes are far apart. Compared to standard classification approaches, this has the advantage that the model can deal with novel classes at test time. Metric learning has several important applications, the most widely studied one being face recognition and face verification.

Throughout the seminar, we will study different approaches to metric learning, mainly based on deep neural networks. We will also discuss different application domains in which metric learning can be applied.