Binary Image Classification: A Genetic Programming Approach to the Problem of Limited Training Instances

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

  author =       "Harith Al-Sahaf and Mengjie Zhang and Mark Johnston",
  title =        "Binary Image Classification: A Genetic Programming
                 Approach to the Problem of Limited Training Instances",
  journal =      "Evolutionary Computation",
  year =         "2016",
  volume =       "24",
  number =       "1",
  pages =        "143--182",
  month =        "Spring",
  keywords =     "genetic algorithms, genetic programming, Local Binary
                 Patterns, One-shot Learning, Image Classification",
  ISSN =         "1063-6560",
  DOI =          "doi:10.1162/EVCO_a_00146",
  size =         "37 pages",
  abstract =     "In the Computer Vision and Pattern Recognition fields,
                 image classification represents an important, yet
                 difficult, task to perform. The remarkable ability of
                 the human visual system, which relies on only one or a
                 few instances to learn a completely new class or an
                 object of a class, is a challenge to build effective
                 computer models to replicate this ability. Recently, we
                 have proposed two Genetic Programming (GP) based
                 methods, One-shot GP and Compound-GP, that aim to
                 evolve a program for the task of binary classification
                 in images. The two methods are designed to use only one
                 or a few instances per class to evolve the model. In
                 this study, we investigate these two methods in terms
                 of performance, robustness, and complexity of the
                 evolved programs. Ten data sets that vary in difficulty
                 have been used to evaluate these two methods. We also
                 compare them with two other GP and six non-GP methods.
                 The results show that One-shot GP and Compound-GP
                 outperform or achieve comparable results to other
                 competitor methods. Moreover, the features extracted by
                 these two methods improve the performance of other
                 classifiers with handcrafted features and those
                 extracted by a recently developed GP-based method in
                 most cases",

Genetic Programming entries for Harith Al-Sahaf Mengjie Zhang Mark Johnston