|
Activity Number:
|
340
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #309894 |
|
Title:
|
A New Clustering Algorithm Based on Self-Updating Process
|
|
Author(s):
|
Ting-Li Chen*+ and Shang-Ying Shiu
|
|
Companies:
|
Academia Sinica
|
|
Address:
|
Institute of Statistical Science, Taipei 115, 115, Taiwan
|
|
Keywords:
|
clustering ; k-means ; self-organizing maps ; image segmentation
|
|
Abstract:
|
Many of the popular clustering methods, such as K-means and Self-Organizing Maps, require a set of initial values to begin the iterative process. In this talk we will present a simple and novel method that does not require such an initial set and can avoid the problem of local minima. The clustering strategy we propose is motivated by intuition on clustering. The algorithm stands from the viewpoint of subjects to be clustered and simulates the process of how they perform self-clustering. At the end of the process subjects belonged to the same cluster would converge to the same point, which represents the cluster location in a p-dimensional space. Our simulation study showed promising results compared to other clustering methods. An example on image segmentation will also be presented.
|
