On Self-Updating Process
- 2013-11-25 (Mon.), 10:30 AM
- Recreation Hall, 2F, Institute of Statistical Science
- Prof. Ting-Li Chen
- Institute of Statistical Science, Academia Sinica
Abstract
Self-Updating Process (SUP) is a clustering algorithm that stands from the viewpoint of data points and simulates the process how data points move and perform self-clustering. This algorithm is particularly effective for data with noise, data with large number of clusters and unbalanced data. In this talk I will discuss the properties of SUP, including the convergence and consistency. A central limit theorem for the converged cluster location by SUP will be presented. Simulation studies show that SUP which blurs the data is more efficient than the corresponding non-blurring method. By combining the concept of gamma-divergence and SUP, the gamma-SUP minimizes the gamma-divergence under the q-Gaussian mixture model and provides a theoretical justification for the clustering results. An application on Cryo-EM images will be presented to demonstrate the performance of gamma-SUP. Another ongoing work that extends SUP from a clustering method to a multi-feature extraction method will also be presented. ??