Summary
International Symposium on Nonlinear Theory and its Applications
2008
Session Number:A1L-E
Session:
Number:A1L-E1
On a New Dissimilarity of Projection Correlation
ENDO Yasunori, UCHIDA Fuyuki, HAMASUNA Yukihiro,
pp.-
Publication Date:2008/9/7
Online ISSN:2188-5079
Summary:
This paper proposes projection correlation which is a new dissimilarity measure for clustering. This dissimilarity has both of the features of cosine correlation and norm-based dissimilarity. The details of the features are discussed by using dissimilarity level curve. Moreover, the effectiveness of the proposed dissimilarity is verified in comparison with the conventional dissimilarities, i.e., squared Euclidean metric and cosine correlation with applying to agglomerative hierarchical clustering.