- author = "Joshua Lee Wilkerson",
- title = "Evolutionary Computing Driven Search Based Software Testing and Correction",
- school = "Computer Science, Missouri University of Science and Technology",
- year = "2012",
- address = "USA",
- keywords = "genetic algorithms, genetic programming, genetic improvement, APR, SBSE, Automatic control, Computer programs, Computer software, Verification, Evolutionary computation, Fault location (Engineering), Program transformation (Computer programming)",
-
URL = "
http://merlin.lib.umsystem.edu/record=b9387992~S5",
-
URL = "http://hdl.handle.net/10355/26508",
-
URL = "
https://mospace.umsystem.edu/xmlui/bitstream/handle/10355/26508/Wilkerson_2012.pdf",
- size = "160 pages",
-
abstract = "For a given program, testing, locating the errors
identified, and correcting those errors is a critical,
yet expensive process. The field of Search Based
Software Engineering (SBSE) addresses these phases by
formulating them as search problems. This dissertation
addresses these challenging problems through the use of
two complimentary evolutionary computing based systems.
The first one is the Fitness Guided Fault Localisation
(FGFL) system, which novelly uses a specification based
fitness function to perform fault localisation. The
second is the Coevolutionary Automated Software
Correction (CASC) system, which employs a variety of
evolutionary computing techniques to perform testing,
correction, and verification of software. In support of
the real world application of these systems, a
practitioner's guide to fitness function design is
provided.
For the FGFL system, experimental results are presented that demonstrate the applicability of fitness guided fault localisation to automate this important phase of software correction in general, and the potential of the FGFL system in particular. For the fitness function design guide, the performance of a guide generated fitness function is compared to that of an expert designed fitness function demonstrating the competitiveness of the guide generated fitness function. For the CASC system, results are presented that demonstrate the system's abilities on a series of problems of both increasing size as well as number of bugs present. The system presented solutions more than 90percent of the time for versions of the programs containing one or two bugs. Additionally, scalability results are presented for the CASC system that indicate that success rate linearly decreases with problem size and that the estimated convergence rate scales at worst linearly with problem size.",
-
notes = "Supervisor: Dr. Daniel Tauritz
Rest of committee Dr. Thomas Weigert Dr. Bruce McMillin Dr. Ali Hurson Dr. Sahra Sedighsarvestani",
