Title :
Application of the HLVQ neural network to hand-written digit recognition
Author :
Solaiman, B. ; Autret, Y.
Author_Institution :
Ecole Nat. Superieure des Telecommun. de Bretagne, Brest, France
Abstract :
In this work, the handwritten digit recognition problem is studied. Self organizing feature maps are mainly considered. The unsupervised Kohonen as well as the hybrid learning vector quantization (HLVQ) algorithms are applied. The main objective is to obtain a topology preserving map having high recognition rates. This is essentially due to the fact that this kind of maps is very useful in realising results interpretations and in the definition of a rejection strategy during the recognition phase
Keywords :
character recognition; self-organising feature maps; topology; unsupervised learning; vector quantisation; handwritten digit recognition; hybrid learning vector quantization; neural network; self organizing feature maps; topology preserving map; unsupervised Kohonen; Bayesian methods; Character recognition; Clustering algorithms; Multi-layer neural network; Network topology; Neural networks; Neurons; Prototypes; Self organizing feature maps; Vector quantization;
Conference_Titel :
Neural Networks for Signal Processing [1994] IV. Proceedings of the 1994 IEEE Workshop
Conference_Location :
Ermioni
Print_ISBN :
0-7803-2026-3
DOI :
10.1109/NNSP.1994.366030
