Stability analysis of self-organizing maps and vector quantization algorithms
Mauro Tucci, Marco Raugi · 2010
In this work a zero input stability analysis of the self-organizing map (SOM) learning algorithm and related unsupervised vector quantization algorithms is presented. The stability of the SOM incremental learning rule is analyzed using the theory of dynamical switched systems. The equation is demonstrated to be asymptotically stable under simple constraints on the learning parameters.