ABSTRACTS OF ARTICLES OF THE JOURNAL "INFORMATION TECHNOLOGIES".
No. 10. Vol. 24. 2018

DOI: 10.17587/it.24.642-648

V. A. Kharakhinov, Graduate Student, e-mail: tes4obse@mail.ru, National Research Irkutsk State Technical University

The Genetic Algorithm as the Alternative Method for Training Kohonen Layer

In this article genetic algorithm suggested as the alternative method for training Kohonen layer. This paper considers evaluations of cluster analysis which were produced by follow methods: K-means; Kohonen layer trained with standard kohonen learning algorithm; Kohonen layer trained with genetic algorithm.
The evaluations of clustering were determined using two indices: the Rand index and Adjusted Rand index.
In addition to this, the article contains the graphical display of the training time for standard kohonen learning algorithm and realized genetic algorithm.
The cluster analysis was accomplished using banknote authentication dataset that described by a set of numerical feature and considers two output flag: authentic and counterfeit.
Keywords: Cluster analysis, K-means, Kohonen layer, genetic algorithm, Rand index

P. 642–648

To the contents