| Project/Area Number |
63550318
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
計測・制御工学
|
|---|
| Research Institution | Science University of Tokyo |
|---|
Principal Investigator |
FUKAO T. Science University of Tokyo, Faculty of Engineering Science, Prof., 基礎工学部, 教授 (90016220)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
WATANABE S. Gunma University, Faculty of Engineering, Associate Prof., 工学部, 助教授 (90008532)
|
|---|
| Project Period (FY) |
1988 – 1989
|
| Project Status |
Completed (Fiscal Year 1989)
|
| Budget Amount *help |
¥2,200,000 (Direct Cost: ¥2,200,000)
Fiscal Year 1989: ¥700,000 (Direct Cost: ¥700,000)
Fiscal Year 1988: ¥1,500,000 (Direct Cost: ¥1,500,000)
|
|---|
| Keywords | distributed and parallel computing / high-speed parallel algorithm / stochastic optimization / global optimization / combinatorial optimization / simulated annealing / network / statistical physical analogy / 統計力学的方法 / 組合わせ論的最適化 / アニーリング法 / 平衡の統計力学的方法 / ネットワーク |
|---|
| Research Abstract |
(1) The collective computations are going to be applied to combinatorial optimization problems. Typical examples are simulated annealing, Boltzmann machine, and neural network. These aim at parallel distributed computing of combinatorial optimization problem expressed by network. Although their principles may be based on the complex and massive connections between network elements(nodes), there may exist a bound of connective complexity in their realizations. To overcome this difficulty we might need some systematizations, such as network in layers, cooperative-competitive mechanism between sub-objectives, etc. We proposed parallel-sequential computation systems(network in layers, parallel networks) due to the decomposition of the objective function in stochastic combinatorial optimization with complicated objective, which reduce massive connections in their realizations. We also confirmed the verification of the method for the TSP(Travelling salesman Problem) by computer experiments.
(2) The simulated annealing is an effective method to solve the global optimization, and we applied it to a complex continuous optimization. Original problem is a design problem of non-linear circuit and it is formulated as an optimization problem with many local minima near the global minimum point. We could get a lot of information from this trial for the application of the simulated annealing to combinatorial optimization problems.
(3) We developed programs for a MIMD multiprocessor system and discussed the problem of centralized-distributed synchronization by computer experiments on the MIMD multiprocessor system, and found some effective schema for parallel algorithms and realization of parallel programs.
|
