If you want to partiticpate, please also enroll in the Learnweb course at https://www.uni-hildesheim.de/learnweb2023/course/view.php?id=2993

Basic knowledge of machine learning (or the simultaneous enrollment in the machine learning lecture) is recommended.

In the seminar, we will study and discuss active machine learning approaches. Active learning is of interest in supervised machine learning scenarios where large amounts of unlabeled data are readily available but labeling data is costly. This scenario matches many application domains, from text classification to image analysis. The main idea behind active learning is that rather than labeling a random subset of all the available unlabeled data, we can try to actively select the most informative data points to label. Typically, this is done in a sequential fashion, by using a classifier trained on the currently available subset of labeled data to estimate how informative potential additional data points might be and then labeling those data points and retraining the classifier.

Throughout the seminar, we will study different methodological approaches to active learning as well as different application domains in which it can be applied.