Analysis of the Superiority of Parameter Optimization over Genetic Programming for a Difficult Object Problem

Created by W.Langdon from gp-bibliography.bib Revision:1.4420

  author =       "Vic Ciesielski and Gayan Wijesinghe and 
                 Andrew Innes and Sabu John",
  title =        "Analysis of the Superiority of Parameter Optimization
                 over Genetic Programming for a Difficult Object
  booktitle =    "Proceedings of the 2006 IEEE Congress on Evolutionary
  year =         "2006",
  editor =       "Gary G. Yen and Lipo Wang and Piero Bonissone and 
                 Simon M. Lucas",
  pages =        "4407--4414",
  address =      "Vancouver",
  month =        "6-21 " # jul,
  publisher =    "IEEE Press",
  keywords =     "genetic algorithms, genetic programming",
  ISBN =         "0-7803-9487-9",
  DOI =          "doi:10.1109/CEC.2006.1688454",
  size =         "8 pages",
  abstract =     "We describe a progression of solutions to a difficult
                 object detection problem, that of locating landmarks in
                 X-Rays used in orthodontic treatment planning. In our
                 first formulation an object detector was a genetic
                 program whose inputs were a number of attributes
                 computed from a scanning window. We used a rich
                 function set comprising + - times divide min; max;
                 ifthenelse. Experimentation with different function
                 sets revealed that using the function set + - gave
                 detectors that were almost as accurate. Such detectors
                 are essentially a linear combination of attributes so
                 we also implemented a parameter optimisation solution
                 with a particle swarm optimiser. Contrary to
                 expectation, the PSO detectors are more accurate and
                 smaller than the GP ones. Our analysis of the reasons
                 for this reveals that (1) the PSO approach involves a
                 considerably smaller search space than the GP approach,
                 (2) in the PSO approach there is a 1-1 mapping between
                 genotype and phenotype while in the GP approach this
                 mapping is many-1 and many semantically equivalent
                 potential solutions are evaluated, (3) the fitness
                 landscape for PSO is a good one for search in that
                 solutions are distributed in areas of high fitness that
                 are easy to locate while the GP landscape is much more
                 difficult to characterise and areas of high fitness
                 more difficult to find.",
  notes =        "WCCI 2006 - A joint meeting of the IEEE, the EPS, and
                 the IEE.

                 IEEE Catalog Number: 06TH8846D IEEE Xplore gives pages
                 = {"}1264--1271{"},",

Genetic Programming entries for Victor Ciesielski Gayan Wijesinghe Andrew Innes Sabu John