Kernel-Based Restricted Boltzmann Machine for Unsupervised Feature Learning 


Vol. 44,  No. 9, pp. 1633-1640, Sep.  2019
10.7840/kics.2019.44.9.1633


PDF Full-Text
  Abstract

This paper presents a restricted Boltzmann machine (RBM) based on kernel method for unsupervised feature learning. The key idea of kernel RBM is that the input data is first implicitly mapped into a high-dimensional feature space, and the energy function is defined with the visible units and hidden units in that space. We propose the use of rectified linear unit for the kernel RBM as a kernel function, which is widely used in deep learning. The gradient-based contrastive divergence algorithm is used for the training of the kernel RBM, and the parameter update rules are derived for the learning. Experimental results on MNIST and STL-10 dataset show that the proposed approach can learn the useful representations, and it outperforms the conventional RBMs on the classification task.

  Statistics
Cumulative Counts from November, 2022
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]

D. K. Kim and J. W. Shin, "Kernel-Based Restricted Boltzmann Machine for Unsupervised Feature Learning," The Journal of Korean Institute of Communications and Information Sciences, vol. 44, no. 9, pp. 1633-1640, 2019. DOI: 10.7840/kics.2019.44.9.1633.

[ACM Style]

Dong Kook Kim and Jong Won Shin. 2019. Kernel-Based Restricted Boltzmann Machine for Unsupervised Feature Learning. The Journal of Korean Institute of Communications and Information Sciences, 44, 9, (2019), 1633-1640. DOI: 10.7840/kics.2019.44.9.1633.

[KICS Style]

Dong Kook Kim and Jong Won Shin, "Kernel-Based Restricted Boltzmann Machine for Unsupervised Feature Learning," The Journal of Korean Institute of Communications and Information Sciences, vol. 44, no. 9, pp. 1633-1640, 9. 2019. (https://doi.org/10.7840/kics.2019.44.9.1633)

Search
(IN TITLE, AUTHOR, ABSTRACT,KEYWORDS)

Advanced Search ▼

POPULAR KEYWORDS
(TOP 10 KEYWORDS)

Recent Publications
(LAST 3 YEARS)






pISSN : 1226 - 4717
eISSN : 2287 - 3880