Abstract
This paper proposes two definition for distance calculation: similarity and link. Similarity can be defined as any formula to
calculate the distance between two vectors, and if the similarity between two points exceeds a threshold, then they are
defined to be neighbors. The link between two vectors are the number of common neighbors between them. A criterion function is
given to calculate the goodness of one cluster. At first, each vector is taken as a cluster, and to merge two clusters, the
one with best goodness is selected and merge it with another one with highest link. This process continues until reach the
number of clusters or the link between every pair of clusters is zero.
Read the Presentation
Slides...
Referred Papers
|