Mechanical a-posteriori Verification of Results: A Case Study for a Safety Critical AI System (2001) [1 citations — 0 self]
Download:
|
|
by Roy Bartsch, Wolfgang Goerigk
In: KHATIB, Lina; PECHEUR, Charles: Model-Based Validation of Intelligence - Papers from 2001 AAAI Spring Symposium. AAAI Press, Menlo Park
http://www.informatik.uni-kiel.de/~wg/New/../Berichte/AAAI-2001.ps.gz
Add To MetaCart
Abstract:
This paper is to show how mechanical theorem proving can be used to verify even complex and heuristic programs like mission critical expert systems. Our approach is mechanical in two ways: The basic idea of runtime result verification is to validate each program result (at runtime) rather than to verify the program itself beforehand. Filtering each result by a sufficient algorithmic correctness predicate guarantees partial correctness of the modified program, if successful checking is proved to imply correctness of the result. We use a mechanical theorem prover to prove the latter fact.
Citations
| 509 | A Computational Logic – Boyer, Moore - 1979 |
| 13 | Program Result Checking Against Adaptive Programs and – Blum, Luby, et al. - 1989 |
| 1 | Mechanisch verifizierte Programmpr ufung fur die Korrektheit von Prufplanen in der Bahntechnik – Bartsch |
