Title :
LVQ-FCV for missing value estimation and pattern classification
Author :
Ikeda, Eriko ; Ichihashi, Hidetomo ; Nagasaka, Kazunori ; Miyoshi, Tetsuya
Author_Institution :
Dept. of Ind. Eng., Osaka Prefecture Univ., Japan
Abstract :
This paper proposes a fuzzy LVQ with prototypes of linear varieties. Minimization of an objective function yields memberships of fuzzy clusters, principal components of the clusters and classification boundaries of LVQ type competitive learning. The proposed fuzzy c-varieties in this paper includes, within the Piccard iteration, a simple procedure for parameter estimation under missing data situations
Keywords :
feedforward neural nets; fuzzy set theory; iterative methods; minimisation; multilayer perceptrons; parameter estimation; pattern classification; principal component analysis; unsupervised learning; vector quantisation; LVQ type competitive learning; LVQ-FCV; PCA; Piccard iteration; classification boundaries; fuzzy LVQ; fuzzy c-varieties; fuzzy cluster memberships; fuzzy multilayer feedforward neural net; missing data situations; missing value estimation; objective function minimization; parameter estimation; pattern classification; principal components; Clustering algorithms; Educational institutions; Industrial engineering; Lagrangian functions; Neurons; Parameter estimation; Pattern classification; Prototypes; Scattering; Vector quantization;
Conference_Titel :
Neural Networks, 1999. IJCNN '99. International Joint Conference on
Conference_Location :
Washington, DC
Print_ISBN :
0-7803-5529-6
DOI :
10.1109/IJCNN.1999.830866
