Title :
Linear fuzzy clustering based on least absolute deviations
Author :
Honda, K. ; Togo, N. ; Ichihashi, H.
Author_Institution :
Osaka Prefecture Univ., Japan
Abstract :
This paper proposes a technique of linear fuzzy clustering based on least absolute deviations. The least absolute deviations adopted in the method provide robust clustering results that are free from the influence of outliers. The simplicity of the proposed objective function makes it possible to handle missing values by simply ignoring only the missing coordinates.
Keywords :
fuzzy set theory; pattern clustering; least absolute deviations; linear fuzzy clustering; robust clustering; Clustering algorithms; Clustering methods; Eigenvalues and eigenfunctions; Frequency estimation; Fuzzy sets; Principal component analysis; Prototypes; Robustness; Scattering; Vectors;
Conference_Titel :
SICE 2002. Proceedings of the 41st SICE Annual Conference
Print_ISBN :
0-7803-7631-5
DOI :
10.1109/SICE.2002.1195770
