Semi-Automated Algorithm Design with Genetic Programming

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

  author =       "Jerry Swan and John R. Woodward",
  title =        "Semi-Automated Algorithm Design with Genetic
  year =         "2015",
  editor =       "Yadahiko Murata",
  address =      "Sendai, Japan",
  month =        "25-28 " # may,
  keywords =     "genetic algorithms, genetic programming",
  URL =          "",
  abstract =     "This tutorial gives a methodology for the use of
                 constructive, generative hyper-heuristics to improve
                 human-designed algorithms. This methodology has been
                 successfully used by the presenters in a variety of
                 application areas. It employs Genetic Programming,
                 resulting in a semi-automated design process. The
                 tutorial provides a simple step-by-step guide,
                 introducing researchers to an exciting topic with many
                 potential applications.

                 In addition to presenting the underlying theory, the
                 tutorial will give an introduction to Templar, a
                 framework that supports Template Method
                 Hyper-heuristics, by which an algorithm is parametrized
                 by a collection of heuristics which are then configured
                 automatically. This method makes effective use of
                 Genetic Programming to tune the algorithm to some
                 target set of problem instances. Since the supporting
                 algorithm skeleton can be arbitrarily complex, this
                 allows greater scalability than can be achieved by the
                 naive application of Genetic Programming. The Templar
                 approach turns the creation of generative
                 hyper-heuristics into the more procedural matter of GP
                 parameter tuning. We describe several case studies,
                 including how to create a hyper-heuristic template for
                 quicksort and show the effectiveness of the approach
                 with a state-of-the-art application in which we
                 optimize quicksort for power consumption. The tutorial
                 is suitable for practitioners at every level, assuming
                 only basic familiarity with metaheuristics and machine
                 learning - some experience of genetic programming is
                 helpful, but not a necessity.",
  notes =        "Tutorial 1000-1145 hrs Mo2-2F4 CEC2015",

Genetic Programming entries for Jerry Swan John R Woodward