• DocumentCode
    885651
  • Title

    A Parapropagation Pattern Classifier

  • Author

    Glucksman, Herbert A.

  • Author_Institution
    Data Sciences Lab., Air Force Cambridge Research Labs. Office of Aerospace Research, U. S. Air Force, Bedford, Mass.
  • Issue
    3
  • fYear
    1965
  • fDate
    6/1/1965 12:00:00 AM
  • Firstpage
    434
  • Lastpage
    443
  • Abstract
    A matrix of simple, identical intercommunicating cells is the core of a pattern classifier that can sort two-valued, two-dimensional patterns, such as alphabetic symbols. The matrix yields information about a patterm mapped on its cells by reporting whether or not a threshold exceeds the number of cells that remain in a ``zero´´ state after a sequence of operations that include directed cell-to-cell propagations of change of state. The pattern furnishes a barrier to the propagations. The testing process is a program written to suit some alphabet of pattern classes. An automatic sequencer for controlling the testing process is described. The testing operations are few, but they can be combined in tests for an abundance of features of great variety. A computer simulation, using a single decision tree, classifies the commonest press and typewriter fonts. A small manually sequenced model employing integrated microelectronic cells has been built.
  • fLanguage
    English
  • Journal_Title
    Electronic Computers, IEEE Transactions on
  • Publisher
    ieee
  • ISSN
    0367-7508
  • Type

    jour

  • DOI
    10.1109/PGEC.1965.264151
  • Filename
    4038463