DESCRIPTION:
Novel use of "soft" correspondence as a data-mining framework to address the problem of matching between different clustering assemblies. The algorithm matches clusters from different clusterings with different weights and computes a correspondence matrix.
ADVANTAGES:
- Improves results of combining multiple clusterings
- Generates intermediate intuitive interpretation of relations between consensus clustering and individual clustering
PATENT STATUS:
Patent Pending (U.S. Patent Application # 11/945,956, Filing Date: 11/27/2007) |
