Very Sparse Random Projection Based Efficient Nystrom Sampling Method Using Coherence Properties 


Vol. 45,  No. 2, pp. 437-443, Feb.  2020
10.7840/kics.2020.45.2.437


PDF
  Abstract

The Nystrom method approximates a low-rank matrix to a Symmetric Positive-Semidefinite (SPSD) matrix by uniformly sampling from its columns. To guarantee an accurate reconstruction, the strong assumption that the chosen column subset matrix has full-rank should be satisfied. It is very difficult to find a sampling method that satisfies these conditions. In this paper, a novel geometrical measure to calculate coherence is utilized indirectly and this term is related to the possibility of acquiring informative columns from the original SPSD matrix. The proposed method uses very sparse random projection to adjust the Coherence value to an appropriate level, which enables successful low-rank matrix approximation of the SPSD matrix. Consequently, comparing with the conventional Nystrom sampling methods, the proposed method is experimentally proved to be efficient in reducing the complexity and close to the best performance method such as one-shot method.

  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]

J. Yoon, J. Lee, Y. Choe, "Very Sparse Random Projection Based Efficient Nystrom Sampling Method Using Coherence Properties," The Journal of Korean Institute of Communications and Information Sciences, vol. 45, no. 2, pp. 437-443, 2020. DOI: 10.7840/kics.2020.45.2.437.

[ACM Style]

Jongsu Yoon, Jieun Lee, and Yoonsik Choe. 2020. Very Sparse Random Projection Based Efficient Nystrom Sampling Method Using Coherence Properties. The Journal of Korean Institute of Communications and Information Sciences, 45, 2, (2020), 437-443. DOI: 10.7840/kics.2020.45.2.437.

[KICS Style]

Jongsu Yoon, Jieun Lee, Yoonsik Choe, "Very Sparse Random Projection Based Efficient Nystrom Sampling Method Using Coherence Properties," The Journal of Korean Institute of Communications and Information Sciences, vol. 45, no. 2, pp. 437-443, 2. 2020. (https://doi.org/10.7840/kics.2020.45.2.437)