The impact of Interconnecting Topologies on SOM Neural Networks

Autor(es)

Pimenta, Mayra Z. ; Comin, Cesar Henrique ; Rodrigues, Francisco A. ; Costa, Luciano da F. ; IEEE

• Tipo de documento: Artigo Científico
• Fonte: 2018 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN); v. N/A, p. 6-pg., 2018-01-01.

Resumo

The effect of different interconnecting topologies on the performance of SOM networks is investigated. Three different databases (digits, musical symbols and letters dataset) are considered, and small world, correlated models, and the deactivation network topologies were taken into account. We show that the topology has a definite influence on the performance of the network at short and middle learning time scales. The addition of long-range connections in the deactivation model tended to reduce accuracy. In addition, we examined the effect of degree-degree correlations, or assortativity, on SOM neural networks. It has been shown that high assortativity values tend to decrease the classification error. (AU)

• Processo FAPESP: 15/18942-8 - Associando Redes Complexas com Espaços Efetivos de Atributos
• Beneficiário: Cesar Henrique Comin
• Modalidade de apoio: Bolsas no Brasil - Pós-Doutorado
• Processo FAPESP: 11/50761-2 - Modelos e métodos de e-Science para ciências da vida e agrárias
• Beneficiário: Roberto Marcondes Cesar Junior
• Modalidade de apoio: Auxílio à Pesquisa - Temático
• Processo FAPESP: 15/22308-2 - Representações intermediárias em Ciência Computacional para descoberta de conhecimento
• Beneficiário: Roberto Marcondes Cesar Junior
• Modalidade de apoio: Auxílio à Pesquisa - Temático