Title :
Certain facts about Kohonen´s LVQ1 algorithm
Author :
Diamantini, Claudia ; Spalvieri, Amaldo
Author_Institution :
Instituto di Inf., Ancona Univ., Italy
fDate :
30 May-2 Jun 1994
Abstract :
LVQ1 is a supervised fine tuning of the Self-Organizing Map, introduced by Kohonen for the classification task. In this paper we present a study about LVQ1. We show that it implements the search of certain saddles on a scalar cost function. The above criterion does not guarantee convergence to a stable solution, as enlightened by a simple example. We also compare the performance of LVQ1 to that of a labeled version of the Self-Organizing Map
Keywords :
learning (artificial intelligence); pattern classification; self-organising feature maps; Kohonen´s LVQ1 algorithm; classification task; labeled version; saddles; scalar cost function; self-organizing map; supervised fine tuning; Algorithm design and analysis; Cost function; Encoding; Error probability; Nearest neighbor searches; Neural networks; Partitioning algorithms; Pattern recognition; Testing; Vectors;
Conference_Titel :
Circuits and Systems, 1994. ISCAS '94., 1994 IEEE International Symposium on
Conference_Location :
London
Print_ISBN :
0-7803-1915-X
DOI :
10.1109/ISCAS.1994.409617
