Discovery of Optical Character Recognition Algorithms using Genetic Programming

Created by W.Langdon from gp-bibliography.bib Revision:1.4504

  author =       "Polina K. Spivak",
  title =        "Discovery of Optical Character Recognition Algorithms
                 using Genetic Programming",
  booktitle =    "Genetic Algorithms and Genetic Programming at Stanford
  year =         "2002",
  editor =       "John R. Koza",
  pages =        "223--232",
  address =      "Stanford, California, 94305-3079 USA",
  month =        jun,
  publisher =    "Stanford Bookstore",
  keywords =     "genetic algorithms, genetic programming",
  URL =          "",
  language =     "en",
  oai =          "oai:CiteSeerXPSU:",
  URL =          "",
  abstract =     "Optical character recognition is a trivial problem, at
                 least for literate humans. However, creating a good
                 character recognition program is not so simple---a
                 single character can have significant variability when
                 considered across many fonts. Some characters are
                 misleadingly similar, especially considering their
                 incarnations in multiple fonts. This paper discusses a
                 technique, as well as some pitfalls, for automatically
                 evolving optical character recognition programs using
                 genetic programming. The problem-specific information
                 required for this technique is a set of training
                 characters, e.g. { 0, {\ldots}, 9, a, {\ldots}, f}, and
                 GIF incarnations of those characters in different
                 fonts. The result is a set of algorithms that can
                 determine which character is represented by an image.
                 The algorithms can be applied to those images used to
                 evolve the algorithms, as well as new images that
                 represent one of the characters in a new font not seen
                 during evolution.",
  notes =        "part of \cite{koza:2002:gagp} 16 characters 10x10 GIF
                 images, 96 fitness cases, parsimony pressure as part of

Genetic Programming entries for Polina K Spivak