Langdon WB, Veerapen N & Ochoa G (2017) Visualising the Search Landscape of the Triangle Program. In: McDermott J, Castelli M, Sekanina L, Haasdijk E & García-Sánchez P (eds.) EuroGP 2017: Genetic Programming. Lecture Notes in Computer Science, 10196. The 20th European Conference on Genetic Programming (EuroGP), Amsterdam, The Netherlands, 19.04.2017-21.04.2017. Cham, Switzerland: Springer, pp. 96-113. https://doi.org/10.1007/978-3-319-55696-3_7
Abstract High order mutation analysis of a software engineering benchmark, including schema and local optima networks, suggests program improvements may not be as hard to find as is often assumed. 1) Bit-wise genetic building blocks are not deceptive and can lead to all global optima. 2) There are many neutral networks, plateaux and local optima, nevertheless in most cases near the human written C source code there are hill climbing routes including neutral moves to solutions.
Keywords genetic improvement; genetic algorithms; genetic programming; software engineering; heuristic methods; test equivalent higher order mutants; fitness landscape; local search