Orthogonal Feature Extraction Scheme for Edge Computing-Based Deep Learning Service 


Vol. 46,  No. 8, pp. 1262-1269, Aug.  2021
10.7840/kics.2021.46.8.1262


PDF
  Abstract

This paper proposes an orthogonal feature extraction scheme in the deep learning service model for end devices with limited computing and networking resources. One way to use the service is to use the service by transmitting the data to be analyzed to a powerful remote server. It may not require a large number of resources for end devices. However, if the size of the transmitted data is large, many resources are needed for communication. Besides, when the remote server is located far from the devices, it is unavailable due to excessive latency. Therefore, it is necessary to reduce the data size without deteriorating the deep learning performance. In the proposed scheme, the device extracts many low-capacity orthogonal feature vectors, and the edge server improves the deep learning performance through an ensemble model using the orthogonal feature vectors. Experimenting on the Fashion-MNIST dataset confirmed that the proposed scheme improves the deep learning service"s performance while reducing the size of transmitted data.

  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. Lee, "Orthogonal Feature Extraction Scheme for Edge Computing-Based Deep Learning Service," The Journal of Korean Institute of Communications and Information Sciences, vol. 46, no. 8, pp. 1262-1269, 2021. DOI: 10.7840/kics.2021.46.8.1262.

[ACM Style]

Jongkwan Lee. 2021. Orthogonal Feature Extraction Scheme for Edge Computing-Based Deep Learning Service. The Journal of Korean Institute of Communications and Information Sciences, 46, 8, (2021), 1262-1269. DOI: 10.7840/kics.2021.46.8.1262.

[KICS Style]

Jongkwan Lee, "Orthogonal Feature Extraction Scheme for Edge Computing-Based Deep Learning Service," The Journal of Korean Institute of Communications and Information Sciences, vol. 46, no. 8, pp. 1262-1269, 8. 2021. (https://doi.org/10.7840/kics.2021.46.8.1262)

Search
(IN TITLE, AUTHOR, ABSTRACT,KEYWORDS)

Advanced Search ▼

POPULAR KEYWORDS
(TOP 10 KEYWORDS)

Recent Publications
(LAST 3 YEARS)






pISSN : 1226 - 4717
eISSN : 2287 - 3880