Skip to main content

BBDC at NIPS

The machine learning group of Klaus Robert Müller is looking back to a successful participation at the NIPS conference. NIPS is among the most influential artificial intelligence conferences. Especially the deep learning community shapes this event.

At this years conference the BBDC members presented a new algorithm to train neural networks. Additionally the group published four workshop papers. Two out of those four received a best paper award! The first describes a novel way to scale Multi-Class-SVM in settings where one discriminates between thousands of different categories. This ability is needed in various Big Data settings. The second paper treats the explanation of non-linear machine learning algorithms. While deep learning, the most known instance of non-linear learning, achieves astonishing results, it is still hardly understood how these algorithms work. Exactly this problem is addressed by the second paper by providing new theoretical and practical ways to explain machine learning models.