B0667
Title: Multiple clustering based Wishart mixture models and its application to seismic data analysis
Authors: Tomoki Tokuda - The University of Tokyo (Japan) [presenting]
Hiromichi Nagao - The University of Tokyo (Japan)
Abstract: For high-dimensional data, it is not straightforward to cluster objects because all features are not always relevant to a particular cluster solution. Some features may be relevant for one cluster solution, but irrelevant for another. In general, in high-dimensional cases, one may assume multiple cluster solutions depending on a specific subset of features. In this situation, a conventional clustering method would not be able to reveal the underlying cluster structure, which is characterized by multiple cluster solutions. So far, effective methods to find such a cluster structure have been less developed. A novel clustering method is discussed, which is useful for revealing the underlying multiple cluster structure. The method is based on Wishart mixture models, which apply to correlation matrices of connectivity data without vectorization. The uniqueness of this method is that multiple cluster solutions are based on particular networks of nodes, optimized in a data-driven manner. Hence, it can identify the underlying pairs of associations between a cluster solution and a node sub-network. The method is applied to seismic data, and it is shown that the method can potentially capture weak tremor, which is otherwise difficult to identify.