Learning from Super-Mutants

Created by W.Langdon from gp-bibliography.bib Revision:1.4524

  author =       "Jason Landsborough and Stephen Harding and 
                 Sunny Fugate",
  title =        "Learning from Super-Mutants",
  booktitle =    "GI-2017",
  year =         "2017",
  editor =       "Justyna Petke and David R. White and W. B. Langdon and 
                 Westley Weimer",
  pages =        "1529--1536",
  address =      "Berlin",
  month =        "15-19 " # jul,
  publisher =    "ACM",
  keywords =     "genetic algorithms, genetic programming, genetic
  isbn13 =       "978-1-4503-4939-0",
  URL =          "http://geneticimprovementofsoftware.com/wp-content/uploads/2017/05/landsborough2017_supermutants.pdf",
  DOI =          "doi:10.1145/3067695.3082525",
  size =         "8 pages",
  abstract =     "In light of recent advances in
                 genetic-algorithm-driven automated program
                 modification, our team has been actively exploring the
                 art, engineering, and discovery of novel
                 semantics-preserving transforms. While modern compilers
                 represent some of the best ideas we have for automated
                 program modification, current approaches represent only
                 a small subset of the types of transforms which can be
                 achieved. In the wilderness of post-apocalyptic
                 software ecosystems of genetically-modified and mutant
                 programs, there exist a broad array of potentially
                 useful software mutations, including
                 semantics-preserving transforms that may play an
                 important role in future software design, development,
                 and most importantly, evolution.",
  notes =        "Mutations (and linear 2pt crossover)to 86x binary
                 machine code generated by GCC (without optimisation).
                 16/107 examples from gnu unix coreutils 8.25. Copy,
                 delete and swap mutation. Sandboxing provided by
                 distributing tests across sixteen 2GB virtual machines
                 each with 2 CPU cores. Half a million fitness
                 evaluations in two-seven days. Blind mutations,
                 probably to code .text section(?) but no information on
                 if mutated code is exercised by test suite or not.
                 SQLite 'our testing approach is sufficient to ensure
                 that a transform is safe in terms of whole-program

                 Section 4.1 Novel Transforms 'transform short circuits
                 the call to close stdout by simply loading the expected
                 return address into rax and leaving stdout open.'
                 'error handling is simply replaced with NOPs'

                 '5 CONCLUSIONS The application of genetic programming
                 to software repair, optimization, and general
                 modification have proven incredibly fruitful.'",

Genetic Programming entries for Jason Landsborough Stephen Harding Sunny Fugate