Title :
Self-organizing maps and SVD
Author_Institution :
Technical University of Ostrava, Czech Republic
Abstract :
Mean square error is usual method to measure quality of self-organizing map (SOM). MSE brings good theoretical description about learning process, but its rather hard to decide what MSE is small enough, or when learning should be stoped. A new technique for measurement of quality of SOM using singular value decomposition will be presented in this paper.
Keywords :
"Self organizing feature maps","Matrix decomposition","Sparse matrices","Singular value decomposition","Neurons","Clustering algorithms","Vector quantization","Databases","Expert systems","Mean square error methods"
Conference_Titel :
Database and Expert Systems Applications, 2007. DEXA ´07. 18th International Workshop on
Print_ISBN :
0-7695-2932-1;978-0-7695-2932-5
DOI :
10.1109/DEXA.2007.86
