| Project/Area Number |
06452398
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
ISHIZUKA Mitsuru Unlv.of Tokyo, Dept.of Eng., Professor, 大学院・工学系研究科, 教授 (50114369)
|
|---|
| Project Period (FY) |
1994 – 1995
|
| Project Status |
Completed (Fiscal Year 1995)
|
| Budget Amount *help |
¥8,200,000 (Direct Cost: ¥8,200,000)
Fiscal Year 1995: ¥2,900,000 (Direct Cost: ¥2,900,000)
Fiscal Year 1994: ¥5,300,000 (Direct Cost: ¥5,300,000)
|
|---|
| Keywords | artificial intelligence / knowledge processing / hypothetical reasoning / fast inference / near-optimal solution / polynomial-time inference / 0-1 integer programming |
|---|
| Research Abstract |
We have achieved a meaningful result in the area of efficient cost-based or weighted) hypothetical reasoning for computing a near-optimal solution which satisfies all given constraints. Prior to this research, we developed a polynomial-time cost-based hypothetical reasoning method, in which the set of described knowledge is transformed into linear inequalities and then an effioient approximate solution method of 0-1 integer programming called pivot and compliment method is applied to compute a near-optimal solution in polynomial time. While this method is very effective, it is hard for us to grasp its behavior operated in mathematical domain ; as a result, we can not improve its efficiency further by considering the knowledge structure of given problems. Thus we have developed a polynomial-time hypothetical reasoning method called networked bubble propagation (NBP) method, which performs an inference operation conceptually similar to the pivot and compliment method on a unique knowledge network. The time-consuming pivoting operation exchanging the states of basis and non-basis nodes (variables in 0-1 integer programming) is improved by considering the knowl-edge structure. The low-order polynomial-time efficiency of the NBP method is shown experimentally. It can be said that the result of this research is also meaningful in bridging symbol-oriented Al reasoning and OR (operation research) computation operated mathematically in multi-dimensional numerical domains.
|
