University of Konstanz
Graduiertenkolleg / PhD Program
Computer and Information Science

Graduation Talks

title

Active Learning in Parallel Universes

speaker

Christian Dietz, University Konstanz
Konstanz, Germany

date & place

Wednesday, 17.07.2013, 15:15 h
Room G 309

abstract

Active learning in parallel universes addresses two challenges at once: learning with a large number of unlabeled and only few labeled objects (active learning) and learning in the presence of multiple descriptor spaces to describe the objects, whereby each of them alone would be insucient for learning a global model (learning in parallel universes). This is a realistic scenario. For example, a common prob- lem in biological time-lapse microscope imaging is the classi cation of cells according to the current phase of the cell-lifecycle. From an active learning point of view, a lot of unlabeled data is available and needs to be labeled to build a robust model. Additionally, cells can be described using various descriptor spaces, the universes (e.g. intensity, Har- alick, Zernike, etc.).