2015 Fiscal Year Final Research Report
Efficient evolutionary algorithms for constrained optimization using landscape modality estimation and rough approximation
Project/Area Number |
24500177
|
Research Category |
Grant-in-Aid for Scientific Research (C)
|
Allocation Type | Multi-year Fund |
Section | 一般 |
Research Field |
Intelligent informatics
|
Research Institution | Hiroshima City University |
Principal Investigator |
|
Project Period (FY) |
2012-04-01 – 2016-03-31
|
Keywords | 最適化アルゴリズム / 制約付き最適化 / 進化的計算 / Differential Evolution / 低精度近似モデル / 近似モデル / 関数形状推定 / ε制約法 |
Outline of Final Research Achievements |
In this study, general-purpose and efficient constrained evolutionary algorithms are studied: (1) We proposed two methods of estimating landscape modality of objective functions by observing the change of objective values in sampling points along a line, or by obtaining the number of valley points which are identified using a proximity graph created from search points. (2) We proposed several tuning methods according to the estimated landscape modality and the methods were applied to differential evolution and particle swarm optimization. (3) We proposed a method that a rough approximation model was applied not only objective functions but also constraint violation functions in order to improve the epsilon constrained method for constrained optimization. It was shown that the efficiency of the constrained optimization algorithms was much improved.
|
Free Research Field |
知能情報学
|