| Project/Area Number |
14580498
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
社会システム工学
|
|---|
| Research Institution | Hiroshima Shudo University |
|---|
Principal Investigator |
TAKAHAMA Setsuko Hiroshima Shudo University, Faculty of Commercial Sciences, Professor, 商学部, 教授 (60186989)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
KODAMA Masanori Hiroshima Shudo University, Faculty of Economic Sciences, Professor, 経済科学部, 教授 (20028989)
HIROMITSU Seijirou Hiroshima Shudo University, Faculty of Economic Sciences, Professor, 経済科学部, 教授 (90043827)
KAIO Naoto Hiroshima Shudo University, Faculty of Economic Sciences, Professor, 経済科学部, 教授 (80148741)
|
|---|
| Project Period (FY) |
2002 – 2004
|
| Project Status |
Completed (Fiscal Year 2004)
|
| Budget Amount *help |
¥3,400,000 (Direct Cost: ¥3,400,000)
Fiscal Year 2004: ¥1,200,000 (Direct Cost: ¥1,200,000)
Fiscal Year 2003: ¥1,100,000 (Direct Cost: ¥1,100,000)
Fiscal Year 2002: ¥1,100,000 (Direct Cost: ¥1,100,000)
|
|---|
| Keywords | multiobjective optimization / constraint optimization problem / structural learning / α constrained method / genetic algorithm / degeneration / coevolution / particle swarm optimizer / RBF / 情報量基準 / 非線形最適化手法 / ニューラルネットワーク |
|---|
| Research Abstract |
In this research, we studied the development and application of optimization methods for nonlinear multiobjective optimization problems in which the differentiability was not guaranteed. The research for three years is approximately classified into the following three groups :
(1)The research on optimization methods for constrained single objective nonlinear optimization problems,
(2)The research on structural learning in which the best model is identified,
(3)The research on structural learning using unconstrained multiobjective optimization methods.
In the category (1), we proposed the transformation method "α constrained method" which converted an algorithm for unconstrained problems into an algorithm for constrained problems. In the a constrained method, the α level comparison which is a lexicographic order is introduced to optimize two objective problem in which the satisfaction level of constraints is maximized and the objective function is minimized simultaneously. Also, we proposed
… More
three methods, the α constrained Simplex method, the a constrained Genetic Algorithm and the a constrained Particle Swarm Optimizer. By comparing these α constrained methods with the standard methods such as GENOCOP5.0, it was shown that these α constrained methods were very efficient methods for solving constrained optimization problems.
In the category (2), we introduced damaged genes into genetic algorithms to realize degeneration and proposed structural learning methods, which efficiently delete the unnecessary parameters in the model, by using the degeneration mechanism. The following were proposed : Genetic Algorithm with Mutant Gene which used the binary coded genes and Genetic Algorithm with Degeneration which used the real coded genes. Also, it was shown that these methods were very effective to optimize the structures of RBF fuzzy control rules and neural networks.
In the category (3), the multiobjective optimization, in which estimation errors and information criteria are both optimized, is introduced to the Genetic Algorithm with Degeneration. It was shown that the way of optimization could achieve more stable structural learning. Less
|
