Development of an adaptive learning type optimization technique based on evolutionary systems and application to VLSI layout design
| Project/Area Number |
18560399
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
System engineering
|
|---|
| Research Institution | Chiba University |
|---|
Principal Investigator |
KOAKUTSU Seiichi Chiba University, Graduate school of Engineering, Associate Professor (70241940)
|
|---|
| Project Period (FY) |
2006 – 2007
|
| Project Status |
Completed (Fiscal Year 2007)
|
| Budget Amount *help |
¥1,940,000 (Direct Cost: ¥1,700,000、Indirect Cost: ¥240,000)
Fiscal Year 2007: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2006: ¥900,000 (Direct Cost: ¥900,000)
|
|---|
| Keywords | Systems Engineering / Soft Computing / Systems Optimization / Placement and Routing Problem |
|---|
| Research Abstract |
This research aims at a development of an adaptive learning type optimization technique that adjusts its search process based on the structure of a cost function. The developed technique is applied to combinatorial optimization problems such as placement and wiring problems of VLSI which have a large-scale and complex cost function. In the developed technique, information on the structure of the cost function is collected in the process of optimization, and it is used to adjust the search process of the solution. The developed technique is effective to quality improvement of the solution and shortening at computation time. The developed technique is based on Genetic Algorithms that imitate biological evolution. In Genetic Algorithms, it is possible to search for the solution of high quality efficiently by encoding the state of the solution in gene, and combining profitable partial solutions by using crossover operators. However, when applying Genetic Algorithms to practical problems, it is not easy to find an ideal gene coding and crossover operators. The feature of the developed technique is to encode the search process in gene, and to introduce the function to optimize the search process according to the feature of the observed cost function into Genetic Algorithms. As a result, the solution of the high quality comes to be obtained efficiently without adjusting the gene coding and the crossover operators of each problem by hand, and without depending on the scale and modeling of problems.
|
Report
(3 results)
Research Products
(10 results)
