We present the now well-known ODE method in a general framework. We state a global convergence theorem when the ODE has no "pseu-docycle" which could apply in many cases. We note that the Kohonen algorithm in dimension 1 (units and stimuli) after self-organization is a cooperative dynamical system and prove the uniqueness of its equilibrium point. We then derive results of convergence (a.s and in distribution) for a very general class of stimuli distributions and neighbourhood functions.
CITATION STYLE
Fort, J. C., & Pagès, G. (1997). Convergences of the Kohonen maps: A dynamical system approach. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 1327, pp. 631–636). Springer Verlag. https://doi.org/10.1007/bfb0020225
Mendeley helps you to discover research relevant for your work.