ECJ.pdf (3.42 MB)

Binary image classification: A genetic programming approach to the problem of limited training instances

Download (3.42 MB)
journal contribution
posted on 27.10.2020, 21:57 by Harith Al-Sahaf, Mengjie Zhang, M Johnston
© 2016 by the Massachusetts Institute of Technology. In the computer vision and pattern recognition fields, image classification represents an important yet difficult task. It is a challenge to build effective computer models to replicate 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. Recently we proposed two genetic programming (GP) 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. We use ten data sets that vary in difficulty 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 results comparable to 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.

History

Preferred citation

Al-Sahaf, H., Zhang, M. & Johnston, M. (2016). Binary image classification: A genetic programming approach to the problem of limited training instances. Evolutionary Computation, 24(1), 143-182. https://doi.org/10.1162/EVCO_a_00146

Journal title

Evolutionary Computation

Volume

24

Issue

1

Publication date

01/01/2016

Pagination

143-182

Publisher

MIT Press - Journals

Publication status

Published

Contribution type

Article

Online publication date

10/03/2016

ISSN

1063-6560

eISSN

1530-9304

Language

en

Exports