A New Algorithm for Optimization of the Kohonen Network Architectures Using the Continuous Hopfield Networks

摘要

The choice of the Kohonen neural network architecture has a great impact on the convergence of trained learning methods. In this paper, we generalize the learning method of the Kohonen network. This method optimizes the Kohonen network architecture and conserves the neighborhood notion defined on the observation set. To this end, we model the problem of Kohonen network architecture optimization on the terms of a mix-integer non linear problem with quadratic constraints. In order to solve the proposed model, we use the nues dynamics method. In this context, the continuous Hopfield network is used in the assignment phase. To show the advantages of our method, some experiments results are introduced.