2002 Fiscal Year Final Research Report Summary
Construction of Program Synthesis System Based on Equivalent Transformation
| Project/Area Number |
12480076
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | HOKKAIDO UNIVERSITY |
|---|
Principal Investigator |
AKAWA Kiyoshi Center for Information and Multimedia Studies, Prof., 情報メディア教育研究総合センター, 教授 (50126265)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
MUNETOMO Masaharu Center for Information and Multimedia Studies, Asso. Prof., 情報メディア教育研究総合センター, 助教授 (00281783)
MIZUTA Masahiro Center for Information and Multimedia Studies, Prof., 情報メディア教育研究総合センター, 教授 (70174026)
|
|---|
| Project Period (FY) |
2000 – 2002
|
| Keywords | Program Synthesis / Meta-Rule / Equivalent Transformation Rule / Definite Clause / Specification / Correctness / Program Transformation / Priority |
|---|
| Research Abstract |
1. Development of a general theory of meta-computation We developed a general theory for program synthesis based on a new class of representation called separated descriptions. We clarified the fundamental difference between program transformation and fule generation, and showed that program transformation can be regarded as a part of our program synthesis.
2. Extension of the program sysnthesis system using first-order expressions We constructed an expertimental program synthesis system, which automatically synthesizes a program by repeatedly generating and accumulating new rules. We obtained several successful results. For example, the system can automatically synthesize, from a definition of a language represented by a first-order expression, a program corresponding to a finite automaton that recognizes the language.
3. Extension of meta-computation We extended meta-decriptions to improve the system to be able to generate a larger class of rules. We introduced expressions of constraints on variables and extended a theory of meta-computation. This enabled the system to generate with applicability conditions and can be more controllable. We also investigated a theoretical foudation of generation of such rules.
4. Evaluation of the program synthesis system We applied the system to generation of parser programs. Highly optimized programs are obtained at the experiments. However they can be more efficient by optimization of execution parts of rules and by transforming the rules into deterministic ones. Although we conducted them manually, complete automation is a future work.
|
