Guest Lecture: Prof. Dr. Amitava Datta, University of Western Australia, Perth. Subspace clustering is an unsupervised machine learning technique. Since high-dimensional datasets are not amenable to clustering in the full-dimensional space (points tend to become almost equally spaced), one way of understanding a dataset is to cluster it in all possible subspaces. Since there are an exponential number (exponential in the dimension) of subspaces, this is very expensive computationally. In fact the best algorithm cannot handle more than 5000-6000 data points and 25 dimensions. We have designed a fast algorithm that can process large datasets of hundreds of dimensions fairly fast. We have even experimented with up to 4000 dimensions and it works.
Public Lecture by Prof. Dr. Amitava Datta
University of Western Australia, Perth.
27 January 2015, 17:30 pm,
Campus Offenburg, Building E, Room E 311
For further information please contact:
<link mail window for sending>Dr. rer. nat. Tobias Lauer