On March 3, 2016, Dr. Krzysztof Dembczynski (Poznan University of Technology, Poland) will give a talk about "Extreme Classification: Machine Learning with extremely large output spaces" in the context of the SFB 901.
Abstract: In many applications of machine learning, such as natural language processing, image classification, speech recognition, the output space can be extremely large, involving hundreds of thousands or even millions of labels. Problems of this kind are challenging from statistical and computational point of view. The questions are whether algorithms can learn in presence of so many labels and scale in time and space to problems of this size. During the talk we will discuss these challenges in depth and present the main existing approaches to meet them.