ALGORITHM OF PRUNING OF HYBRID NEURAL NETWORKS ENSEMBLE

Authors

  • O. I. Chumachenko National Technical University of Ukraine “Ihor Sikorsky Kyiv Polytechnic Institute,” Kyiv, Ukraine
  • A. O. Kuzmenko National Aviation University, Kyiv, Ukraine

DOI:

https://doi.org/10.18372/1990-5548.55.12772

Keywords:

Ensemble pruning, bagging, accuracy, diversity

Abstract

Despite the fact that the ensemble is usually more accurate than a single network, existing ensemble techniques tend to create unreasonably large ensembles that increase the use of memory and computation costs. The ensemble's pruning solves this problem. The article analyzes the compromise between accuracy and diversity and it is proved that classifiers, which are more accurate and make more predictions in the minority group, are more important for the construction of the subensemble. A metric that takes into account accuracy and diversity is proposed to evaluate the contribution of a separate classifier that will help to allocate the required number of networks with the best results.

Author Biographies

O. I. Chumachenko, National Technical University of Ukraine “Ihor Sikorsky Kyiv Polytechnic Institute,” Kyiv, Ukraine

Technical Cybernetic Department

Candidate of Science (Engineering). Assosiate Professor.

A. O. Kuzmenko, National Aviation University, Kyiv, Ukraine

Aviation Computer-Integrated Complexes Department

6th year of study student

References

Zhenyu Lu, Xindong Wu, Xingquan Zhu, and Josh Bongard, Ensemble Pruning via Individual Contribution Ordering. Burlington. 2007, 10 p.

Gonzalo Martı´nez-Munoz, Daniel Herna´ndez-Lobato, and Alberto Sua´rez, “An Analysis of Ensemble Pruning Techniques Based on Ordered Aggregation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, no. 2, 2009, pp. 245–259.

Z.-H. Zhou, J. Wu, and W. Tang, “Ensembling Neural Networks: Many Could Be Better than All,” Artificial Intelligence, vol. 137, pp. 239–263, 2002.

Downloads

Issue

Vol. 1 No. 55 (2018)

Section

COMPUTER-AIDED DESIGN SYSTEMS

License

Authors who publish with this journal agree to the following terms:

Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.

Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.

Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See The Effect of Open Access).