Improving evolutionary algorithms from population structures and interaction networks

2018 Fiscal Year Final Research Report

• Project/Area Number: 17K12751
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Soft computing
• Research Institution: University of Toyama
• Principal Investigator: Gao Shangce 富山大学, 大学院理工学研究部(工学), 准教授 (60734572)
• Project Period (FY): 2017-04-01 – 2019-03-31
• Keywords: 進化計算 / アルゴリズム / 最適化

Outline of Final Research Achievements

In this research, we systematically study the influence of the population structure for the evolutionary algorithms (EA). We construct a novel information interaction network of the population (PIN) to extract the characteristics of the search dynamics in EA, and then analyze the design of effective search algorithm from the perspective of structure. The distribution of node degree in PIN is introduced, based on which several algorithms (e.g., differential evolution) are improved in terms of solution accuracy and population diversity. Furthermore, the performance of improved algorithms are verified by applying them on various problems, such as multivalued network learning, dendritic neuron learning, protein structure prediction, and so on.

Free Research Field

計算知能

Academic Significance and Societal Importance of the Research Achievements

本研究の成果について、学術的にみると、進化計算の探索性能と集団構造との間のいくつかの一般的な関係を引き出すことができ、それによってより性能の良い進化計算の設計や進化計算の改善に対するいくつかの潜在的なガイドラインを与えることができると予想されます。特に、異種集団構造に基づいて構築された進化計算は、より効果的であると期待できます。また、社会的にみると、進化計算のコミュニティ内で大きな影響を及ぼし、さらに多目的または動的な環境下で実世界の問題を解決するためのより強力な計算手法を提供できます。