A Sequential Algorithm for Training the SOM Prototypes Based on Higher-Order Recursive Equations

A n ovel t r aining algor i thm i s proposed for the for mation of Self-Organizing Maps (SOM). In the proposed m odel, t he weig htsare update d incrementally by using a hig her- order d ifference equation, w hich i mplements a low-pass dig ital filter. It is possible toimprove s electe d features of the s elf-organization pro cess w i th re spect to t he basic SOM by suitably desig ning the fi lter. Moreover,from this model, new v isualization to ols can b e der ived for clu ster v i sualization a nd for m onitor ing the qualit y of the m ap.