Title :
Determination of initial configuration for LVQ by using CNN
Author :
Kim, Baek-Sop ; Lee, Sang Hee ; Kim, Dae Keuk
Author_Institution :
Dept. of Comput. Sci., Hallym Univ., Chunchon, South Korea
Abstract :
A method for determining the initial configuration for the LVQ is proposed. It is based on the condensed nearest neighbor (CNN) rule followed by the K-means clustering method. Experiments show that the proposed method is generally better than the conventional ones which use k-NN or the K-means. And it is also shown that the performance of the CNN is improved by applying the LVQ as a post processing.
Keywords :
learning (artificial intelligence); neural nets; pattern recognition; vector quantisation; CNN; K-means clustering method; LVQ; condensed nearest neighbor rule; initial configuration; post processing; Bayesian methods; Cellular neural networks; Clustering algorithms; Clustering methods; Computer science; Error analysis; Nearest neighbor searches; Neural networks; Pattern recognition; Speech recognition;
Conference_Titel :
Neural Networks, 1993. IJCNN '93-Nagoya. Proceedings of 1993 International Joint Conference on
Print_ISBN :
0-7803-1421-2
DOI :
10.1109/IJCNN.1993.714221
