REK-SVM: a robust and efficient SVM algorithm based on K-medians clustering

Chongjun Gao; Nong Sang; Jiahui Lei

doi:10.1117/12.2524251

6 May 2019 REK-SVM: a robust and efficient SVM algorithm based on K-medians clustering

Chongjun Gao, Nong Sang, Jiahui Lei

Proceedings Volume 11069, Tenth International Conference on Graphics and Image Processing (ICGIP 2018); 1106907 (2019) https://doi.org/10.1117/12.2524251
Event: Tenth International Conference on Graphic and Image Processing (ICGIP 2018), 2018, Chengdu, China

Abstract

Support vector machines (SVMs) have been widely used for binary classification. But large-scale training set will bring huge computation to the SVM. Researcher have proposed many techniques to improve the training efficiency of SVMs, and a typical class of improved SVMs is based on sparsely reducing training samples. To achieve this, clustering-based methods are most commonly used. However, clustering-based methods are ready to be disturbed by noise points. In order to solve this problem, this paper proposes a robust and efficient SVM algorithm based on K-Medians clustering (REK-SVM). Here, for each cluster, the cluster center takes the median value of each dimension attribute in the cluster, which can reduce the noise points. Especially, when the number of noise points distributed discretely is less than half of the total number of samples in the cluster, noise interference can be completely removed. The noise-free or noise-reduced subset data is used to train the SVM model. Experimental results show that our algorithm is fast and effective. For the processing of noise-containing classification data, its performance far exceeds SVM in terms of classification accuracy and efficiency. Compared to the K-SVM, they have the same computational complexity, but our algorithm is much higher than K-SVM in classification accuracy.

Citation Download Citation

Chongjun Gao, Nong Sang, and Jiahui Lei "REK-SVM: a robust and efficient SVM algorithm based on K-medians clustering", Proc. SPIE 11069, Tenth International Conference on Graphics and Image Processing (ICGIP 2018), 1106907 (6 May 2019); https://doi.org/10.1117/12.2524251

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available