Improving the Training Performance of Support Vector Machines Using Momentum


The Transactions of the Korea Information Processing Society (1994 ~ 2000), Vol. 7, No. 5, pp. 1446-1455, May. 2000
10.3745/KIPSTE.2000.7.5.1446,   PDF Download:

Abstract

This paper proposes a modified kernel-adatron algorithm which is a learning algorithm for the support vector machines. It is added the gradient ascent to a fraction of previous solution-change value, momentum. The momentum is applied for increasing the speed of convergence by suppressing the oscillations as the optimal solution converges. It can achieve a superior property of the implementations in the kernel-adatron algorithm. The SVM using the proposed algorithm has been applied to classify the nonlinear problems with 18 data and 200 breast cancer databases by 2-class. The simulation results shows that the proposed algorithm has better performances of the learning time and the classification for test data, in comparison with those using the conventional QP and Campbell's kernel-adatron algorithm.


Statistics
Show / Hide Statistics

Statistics (Cumulative Counts from September 1st, 2017)
Multiple requests among the same browser session are counted as one view.
If you mouse over a chart, the values of data points will be shown.


Cite this article
[IEEE Style]
Y. H. Cho, "Improving the Training Performance of Support Vector Machines Using Momentum," The Transactions of the Korea Information Processing Society (1994 ~ 2000), vol. 7, no. 5, pp. 1446-1455, 2000. DOI: 10.3745/KIPSTE.2000.7.5.1446.

[ACM Style]
Yong Hyun Cho. 2000. Improving the Training Performance of Support Vector Machines Using Momentum. The Transactions of the Korea Information Processing Society (1994 ~ 2000), 7, 5, (2000), 1446-1455. DOI: 10.3745/KIPSTE.2000.7.5.1446.