TU Clausthal, Germany
Clustering constitutes a fundamental problem in information processing for
data mining, data compression, data visualization, data preprocessing etc.
A variety of models exists which are based on different principles, such as model
based clustering, distance based algorithms, or hierarchical versions.
In the talk, we will focus on recent extensions of intuitive prototype based
clustering models to powerful alternatives which address the following issues:
(i) Insensitivity with respect to initialization by means of neighborhood incorporation
(ii) Clustering and visualization incorporating auxiliary information given by class labels
(iii) Clustering noneuclidean data based on a general distance matrix
(iv) Clustering and visualization of time series by means of recursive models.
The effectivity of these models will be demonstrated on a variety of problems
from pattern recognition and computational biology.
back to the list of talks