Title :
Improvements to Kohonen self-organising algorithm
Author_Institution :
Dept. of Phys., Keele Univ.
fDate :
3/17/1994 12:00:00 AM
Abstract :
Two simple methods are proposed to speed up the Kohonen selforganising algorithm on conventional computers. One method is called partial distance elimination and the other method is based on the topology preservation properties of a Kohonen map and statistical correlation of input vectors. Computer simulation shows that for real speech data more than 50% of CPU time can be saved
Keywords :
parallel algorithms; self-organising feature maps; speech analysis and processing; CPU time; Kohonen self-organising algorithm; computer simulation; input vectors; partial distance elimination; speech data; statistical correlation; topology preservation properties;
Journal_Title :
Electronics Letters
DOI :
10.1049/el:19940346
