Guiding Evolutionary Learning by Searching for Regularities in Behavioral Trajectories: A Case for Representation Agnosticism

Created by W.Langdon from gp-bibliography.bib Revision:1.4496

  author =       "Krzysztof Krawiec and Jerry Swan",
  title =        "Guiding Evolutionary Learning by Searching for
                 Regularities in Behavioral Trajectories: A Case for
                 Representation Agnosticism",
  booktitle =    "How Should Intelligence Be Abstracted in AI Research:
                 {MDPs}, Symbolic Representations, Artificial Neural
                 Networks, or ...",
  year =         "2013",
  editor =       "Sebastian Risi and Joel Lehman and Jeff Clune",
  number =       "FS-13-02",
  series =       "2013 AAAI Fall Symposium Series",
  pages =        "41--46",
  address =      "Arlington, Virginia, USA",
  month =        "15-17 " # nov,
  organisation = "Association for the Advancement of Artificial
  publisher =    "AAAI Press",
  publisher_address = "Menlo Park, California, USA",
  keywords =     "genetic algorithms, genetic programming, pattern
                 guided GP",
  isbn13 =       "978-1-57735-612-7",
  URL =          "",
  URL =          "",
  size =         "6 pages",
  abstract =     "An intelligent agent can display behaviour that is not
                 directly related to the task it learns. Depending on
                 the adopted AI framework and task formulation, such
                 behaviour is sometimes attributed to environment
                 exploration, or ignored as irrelevant, or even
                 penalised as undesired. We postulate here that
                 virtually every interaction of an agent with its
                 learning environment can result in outcomes that carry
                 information which can be potentially exploited to solve
                 the task. To support this claim, we present Pattern
                 Guided Evolutionary Algorithm (PANGEA), an extension of
                 genetic programming (GP), a genre of evolutionary
                 computation that aims at synthesising programs that
                 display the desired input-output behavior. PANGEA uses
                 machine learning to search for regularities in
                 intermediate outcomes of program execution (which are
                 ignored in standard GP), more specifically for
                 relationships between these outcomes and the desired
                 program output. The information elicited in this way is
                 used to guide the evolutionary learning process by
                 appropriately adjusting program fitness. An experiment
                 conducted on a suite of benchmarks demonstrates that
                 this architecture makes agent learning more effective
                 than in conventional GP. In the paper, we discuss the
                 possible generalisations and extensions of this
                 architecture and its relationships with other
                 contemporary paradigms like novelty search and deep
                 learning. In conclusion, we extrapolate PANGEA to
                 postulate a dynamic and behavioural learning framework
                 for intelligent agents.",
  notes =        "",

Genetic Programming entries for Krzysztof Krawiec Jerry Swan