Octasom an octagonal based som lattice structure for biomedical problems

Abstract

In this study, an octagonal-based self-organizing network’s lattice structure is proposed to allow more exploration and exploitation in updating the weights for better mapping and classification performances. The neighborhood of the octagonal-based lattice structure provides more nodes for the weights updating than standard hexagonal-based lattice structure. Based on our experiment, the octagonal-based lattice structure performance is better than standard hexagonal lattice structure on biomedical datasets for classification problem. This indicates that proposed algorithm is an alternative lattice structure for self-organizing network which give more wisdom to classification problems especially in the biomedical domains.