The course includes the following two sections.
Section: Natural Language Processing
Prof. Paola Velardi
March 2nd -April 13th
Programme
Section: Dimensionality Reduction and Hashing for Machine Learning and Information Retrieval
Prof. Andreas Karwath, Univ. Mainz, Germany
April 20th - May 25th
Programme
In the seminar we will present a number of techniques designed to reduce the dimensions of high-dimensional datasets. Those datasets occur frequently in many real-word applications such as document retrieval, the semantic web, or in image similarity searches. The key idea of the techniques presented, is to reduce the number of features describing each data point, either by selecting the most informative features (feature selection) or by constructing novel features (feature construction). Within the seminar we will investigate both approaches, the main part however will be about novel automatic feature construction. The presented techniques include PCA, MDS, LLE, Isomap, Semantic Hashing, as well as Spectral Hashing as well as applications to use these techniques for prediction labels of object within point clouds. Lecture plan:
Exams
The exam includes two parts and will require paper presentations and/or a practical assignment (only in one of the two sections).
Book through Infostud (please email the coordinator in case there is no registration open). The next exam dates are:
June 2012