Science, technology, and commerce increasingly recognize the importance of machine learning approaches for data-intensive, evidence based decision making.

While the number of machine learning applications and the volume of data increases, resources like the capacities of processing systems or human supervisors remain limited. This makes active learning techniques an important but challenging research topic. Active Learning bridges the gap between data-centric and user-centric approaches by optimizing their interaction, e.g., by selecting the most relevant information, by performing the most informative experiment, or by selecting solely the most informative data for processing. Thereby, it enables efficient allocation of limited resources, thus reducing costs in terms of time (e.g., human effort or processing time) and money.

Active Learning is a very useful methodology in on-line industrial applications to minimize the effort for sample annotation and measurements of "target" values (e.g., quality criteria). It further reduces the computation load of machine learning and data mining tools, as embedded models are only updated based on a subset of samples selected by the implemented active learning technique. Especially, in cost-intensive areas like medical applications (e.g., diagnostic support, brain-computer interfaces) the efficient use of expert knowledge is crucial.

However, there are several recent research directions, open problems, and challenges in active learning, which ideally should be addressed and discussed in this workshop.