search for


Comparison of nonlinear classification methods for image data
Journal of the Korean Data & Information Science Society 2021;32:767-80
Published online July 31, 2021;
© 2021 Korean Data and Information Science Society.

Kyuri Park1 · Changyi Park2

12Department of Statistics, University of Seoul
Correspondence to: This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT) (No. 2020R1F1A01048268).
1 Graduate student, Department of Statistics, University of Seoul, Seoul 02504, Korea.
2 Professor, Department of Statistics, University of Seoul, Seoul 02504, Korea.
Received June 11, 2021; Revised June 30, 2021; Accepted July 2, 2021.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License ( which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Image classification is one of the most actively studied topics in machine learning. Image data generally has a two-dimensional or three-dimensional matrix structure, and vectorization is performed to apply traditional classification techniques such as support vector machine (SVM). However, vectorization may ignore the structural information provided by image data. Convolutional neural network (CNN) using structural information has been introduced as a remedy to the drawback, but neural networks including CNN generally require a lot of data. On the other hand, SVM shows stable classification performances even with a small number of samples, and extensions of SVM reflecting structural information such as support matrix machine (SMM) and kernel support matrix machine (KSMM) have been recently proposed. In this paper, we compare the predictive performances of SVM, SMM, KSMM, and CNN on image data with relatively small number of samples.
Keywords : Convolutional neural network, kernel, support matrix machine, support vector machine.