Numerical algorithms for graph problems and its parallelization
| Project/Area Number |
26600155
|
|---|
| Research Category |
Grant-in-Aid for Challenging Exploratory Research
|
| Allocation Type | Multi-year Fund |
|---|
| Research Field |
Computational science
|
|---|
| Research Institution | Chuo University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2014-04-01 – 2018-03-31
|
| Project Status |
Completed (Fiscal Year 2017)
|
| Budget Amount *help |
¥3,120,000 (Direct Cost: ¥2,400,000、Indirect Cost: ¥720,000)
Fiscal Year 2016: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Fiscal Year 2015: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2014: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
|
|---|
| Keywords | 自動微分 / グラフの数値処理 / 高階微分 / アルゴリズム自動微分 / 区分的微分可能関数 |
|---|
| Outline of Final Research Achievements |
Numerical parallel approach to graph problems was developed and investigated. Ordinal numerical parallel techniques could be utilized for solving graph problems with our approach, and the usefulness of numerical parallel computation were confirmed with numerical experiments with 100 Linux personal computers with GPU. Developing the new algorithm for computing the exact value by an approximate value of the summation is still continuing and not completed yet.
In the field of algorithmic differentiation that is the foundation of this research, a new technique for processing the absolute operations was proposed by Griewank with a new formalization called ABS(Absolute)-normal-form. Since he visited us at October 2015, studying on this new formalization that was related to our approach, we developed a new enumeration algorithm of subdifferentials of piecewise linearization. Presenting this research in two workshops, we published a paper that was accepted by a journal in this April.
|
Report
(5 results)
Research Products
(6 results)
