Development of self-evolving modular network which grows to the optimal structure in self-organizing manner

2015 Fiscal Year Final Research Report

• Project/Area Number: 25730153
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Soft computing
• Research Institution: National Fisheries University
• Principal Investigator: Tokunaga Kazuhiro 独立行政法人水産大学校, その他部局等, 講師 (00432956)
• Project Period (FY): 2013-04-01 – 2016-03-31
• Keywords: 自己成長型モジュラーネットワーク / 成長型自己組織化マップ

Outline of Final Research Achievements

The purpose of this work is to develop new growing neural network for the backbone algorithm of a self-evolving modular network. In this work, an information criterion-based growing topology representing has been proposed. Also its theory and learning algorithm have been established. In the experiments with artificial data, the proposed method has been able to obtain a sufficient results with the less samples as compared with the conventional growing neural network. Also the stability of the learning is shown in the proposed method. From the results of various experiments, it was suggested that the proposed method is effective as a backbone algorithm of the self-evolving modular network.

Free Research Field

脳情報工学