- author = "Aymeric Blot and Justyna Petke",
- title = "Comparing Genetic Programming Approaches for Non-Functional Genetic Improvement Case Study: Improvement of {MiniSAT's} Running Time",
- booktitle = "EuroGP 2020: Proceedings of the 23rd European Conference on Genetic Programming",
- year = "2020",
- month = "15-17 " # apr,
- editor = "Ting Hu and Nuno Lourenco and Eric Medvet",
- series = "LNCS",
- volume = "12101",
- publisher = "Springer Verlag",
- address = "Seville, Spain",
- pages = "68--83",
- organisation = "EvoStar, Species",
- keywords = "genetic algorithms, genetic programming, Genetic improvement, GI, GP, Search based software engineering, SBSE, Boolean satisfiability, SAT, minisat-2.2.0, linux perf, AST, XML",
- isbn13 = "978-3-030-44093-0",
-
URL = "
http://www.cs.ucl.ac.uk/staff/a.blot/files/blot_eurogp_2020.pdf",
-
URL = "
https://discovery.ucl.ac.uk/id/eprint/10097404",
-
video_url = "https://www.youtube.com/watch?v=D-UQr-P3zUQ",
-
DOI = "
doi:10.1007/978-3-030-44094-7_5",
- abstract = "Genetic improvement (GI) uses automated search to findimproved versions of existing software. While most GI work use geneticprogramming (GP) as the underlying search process, focus is usuallygiven to the target software only. As a result, specifics of GP algorithmsfor GI are not well understood and rarely compared to one another. In this work, we propose a robust experimental protocol to compare different GI search processes and investigate several variants of GP and random-based approaches. Through repeated experiments, we report acomparative analysis of these approaches, using one of the previously used GI scenarios: improvement of runtime of the MiniSAT satisfiability solver. We conclude that the test suites used have the most significant impact on the GI results. Both random and GP-based approaches are able to find improved software, even though the percentage of viable software variants is significantly smaller in the random case (14.5percent vs. 80.1percent). We also report that GI produces MiniSAT variants up to twice as fast as the original on sets of previously unseen instances from the same application domain.",
-
notes = "slides
http://www.cs.ucl.ac.uk/staff/a.blot/files/blot_eurogp_2020_slides.pdf
PyGGI https://github.com/coinse/pyggi
minisat 2. 80.1pecent of mutants GP compile, run, produce right answer. About 70percent faster than original, based on instruction count (slide 13). But... (slide 14) over fitting. Hetrogenous SAT dataset: very easy to hard. video 20 mins + questions
Figure 2: 5-fold cross validation. CentOS-7 Linux kernel 3.10.0 and GCC 4.8.5. 130 combinatorial interaction testing (CIT) instances.
SrcML AST https://www.srcml.org/ also known as \cite{DBLP:conf/eurogp/BlotP20}
http://www.evostar.org/2020/cfp_eurogp.php Part of \cite{Hu:2020:GP} EuroGP'2020 held in conjunction with EvoCOP2020, EvoMusArt2020 and EvoApplications2020",
