2022 Fiscal Year Final Research Report
Improvement of solving speed for nonlinear optimization problems by avoiding the deterioration of numerical condition outside the neighborhod of a optimal solution
Project/Area Number |
18K11185
|
Research Category |
Grant-in-Aid for Scientific Research (C)
|
Allocation Type | Multi-year Fund |
Section | 一般 |
Review Section |
Basic Section 60020:Mathematical informatics-related
|
Research Institution | Kansai University |
Principal Investigator |
Dan Hiroshige 関西大学, 環境都市工学部, 教授 (30434822)
|
Project Period (FY) |
2018-04-01 – 2023-03-31
|
Keywords | 非線形最適化問題 / 最適化ソフトウェア / 自動微分 / モデリング言語 |
Outline of Final Research Achievements |
In this study, we have implemented an automatic differentiation library for large-scale nonlinear optimization problems and a library for solving nonlinear optimization problems. In general, when solving nonlinear optimization problems, automatic differentiation is used to compute partial derivative values of the functions that compose the problem. On the other hand, in large-scale nonlinear optimization problems, functions appearing in the problem are often indexed and have the same structure. In this study, we have implemented a library that can handle functions with the same structure in a fast manner. In addition, we have developed a comprehensive library for solving nonlinear optimization problems. This library can be used to implement new algorithms for solving nonlinear optimization problems in the future.
|
Free Research Field |
数理最適化
|
Academic Significance and Societal Importance of the Research Achievements |
非線形最適化問題に関する研究は,求解アルゴリズムの大域的な収束性や,最適解近傍での収束速度に関するものなど,理論的なものが多い.しかし本来,非線形最適化問題は,アルゴリズムの理論的な研究と,問題を実際に解くためのソフトウェアの整備が両輪となることで,そのポテンシャルを発揮することができる. 本研究では,非線形最適化問題を求解するためのライブラリを包括的に実装することに焦点を当てた研究を行った.具体的には,求解アルゴリズムそのもの,あるいはその入出力,さらには複数のアルゴリズムで共通する処理を実装した.これにより,実際に非線形最適化を解いたり,新たな求解アルゴリズムの実装に役立てることができる.
|