2021 Fiscal Year Final Research Report
Proposal of new evaluation value for combinatorial optimization problem using deep learning
| Project/Area Number |
18K11484
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 61040:Soft computing-related
|
|---|
| Research Institution | Kansai University |
|---|
Principal Investigator |
EBARA Hiroyuki 関西大学, システム理工学部, 教授 (50194014)
|
|---|
| Project Period (FY) |
2018-04-01 – 2022-03-31
|
| Keywords | 深層学習 / 強化学習 / 機械学習 / 組合せ最適化問題 / 巡回セールスマン問題 |
|---|
| Outline of Final Research Achievements |
In this research, I propose approximate solution methods that apply deep learning and deep reinforcement learning to the traveling salesman problem, which is one of the combinatorial optimization problems. The deep learning is trained to learn the optimum solution or the near-optimal solution as an image for a large number of randomly generated problem instances, and an evaluation value is calculated from the image data obtained as an output. The evaluation value obtained by learning is adopted to the heuristic solution method instead of the conventional evaluation value based on distance, in order to obtain a solution. Computational experiments have shown that the evaluation values obtained by learning are valid.
|
| Free Research Field |
離散最適化
|
| Academic Significance and Societal Importance of the Research Achievements |
組合せ最適化問題、特に、巡回セールスマン問題は応用範囲がたいへん広く、宅配便の配送計画や、ロボットの動作計画、プリント基板のドリル穴空け問題などに応用できるため、社会的意義は大きいと考える。
さらに、巡回セールスマン問題は、最も研究されている組合せ最適化問題の1つであり、この問題にディープラーニング手法の有効性を示すことは、他の組合せ最適化問題へも応用を示唆するとともに、ディープラーニングの新たな応用分野を示すことにもなっており、学術的意義は大きいと考える。
|
