Fitness Landscape Learning Evolutionary Computation by means of Machine Learning

2018 Fiscal Year Final Research Report

• Project/Area Number: 26330282
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Soft computing
• Research Institution: Osaka Prefecture University
• Principal Investigator: Mori Naoki 大阪府立大学, 工学(系)研究科(研究院), 准教授 (90295717)
• Project Period (FY): 2014-04-01 – 2019-03-31
• Keywords: 進化型計算 / 機械学習 / 深層学習 / 適応度景観推定型進化型計算

Outline of Final Research Achievements

In this research, I proposed the Fitness Landscape Learning Evolution Computation (FLLEC) which can estimate the fitness landscape by means of machine learning techniques with the background of the remarkable improvement of machine learning in recent years.
We focused on combining deep learning and evolutionary computing, and applying the proposed method to the real problem. I have gotten lots of results by FLLEC such as understanding of human creation especially four-scene comics, the agent strategy evolution of the real stock market, and automatic music generation based on Genetic programming and Variational AutoEncoder (VAE). As a result, I showed that it is possible to construct excellent artificial intelligence by combining machine learning and evolutionary computation.

Free Research Field

進化型計算

Academic Significance and Societal Importance of the Research Achievements

近年，計算機による問題解決は人工知能の発展を背景として複雑さを増し続けている．そこで，本研究では複雑な問題が持つ適応度景観を推定することで適応的に問題を解くことが可能な適応度景観推定型進化型計算(FLLEC) を提案した．特にFLLECに近年急速に発展している機械学習手法を導入することで大幅な性能の向上を実現した．また，人の創作物理解, 株取引，音楽の自動生成など実問題への適用に関する数値実験を通して提案手法の有効性を具体的に示した．
結果として機械学習と進化型計算の融合によって優れた人工知能の構築が可能であることを示した．