180 likes | 306 Views
Topology-Based Hierarchical Clustering of Self-Organizing Maps. Presenter : Wu, Min-Cong Authors : Kadim Ta¸sdemir , Pavel Milenov , and Brooke Tapsall 2011,IEEE. Outlines. Motivation Objectives Methodology Experiments Conclusions Comments. Motivation.
E N D
Topology-Based Hierarchical Clustering of Self-Organizing Maps Presenter : Wu, Min-Cong Authors : KadimTa¸sdemir, Pavel Milenov, and Brooke Tapsall2011,IEEE
Outlines • Motivation • Objectives • Methodology • Experiments • Conclusions • Comments
Motivation • Hierarchical clustering various distance-based similarity measures that have some flaw . • 1. sensitivity to inhomogeneous within-cluster density distributions, noise or outliers. • 2. depend on the cluster centroids and dispersion around these centroids.
Objectives • we employ average linkage for hierarchical clustering of prototypes based on CONN so that at each agglomeration step we merge the pair with maximum average between cluster connectivity that method CONN linkage, and add a new similarity criteria CONN_Index.
Methodology-SOMs • Adapted BMU • Updating BMU • RFi and RFij
Methodology-Connectivity Martix from CONN CONN CADJ CONN(P1,P2) = 3 CONN(P1,P3) = 5 CONN(P2,P3) = 2 CONN(P3,P2) = 2 CONN(P3,P1) = 5 CONN(P2,P1) = 3 CONN=CADJ(p1,p2)+CADJ(p2,p1) =1+2 =3
Methodology-CONN Linkage Similary matrix Delete Add
Methodology-Applicability and Complexity of the Algorithm represent the data topology occasionally Delaunay graph is to have dense enough prototypes. CONN Linkage’s time complexity = O(p^2*d) Average Linkage’s time complexity = O(p^3*d)
Conclusions • CONN linkage produces partitionings better than the ones obtained by distance-based linkages. • Conn_Index based on CONN graph provided better decisions than other indices in the study reported in this paper.
Comments • Advantages • CONN_Index provides the best partitioning for the preset number of clusters. • CONN linkage is mainly proposed for accurate clustering of remote sensing imagery • Applications • hierarchical clustering.