Convolution with Logarithmic Filter Groups for Efficient Shallow CNN

Tae Kwan Lee; Wissam J. Baddar; Seong Tae Kim; Yong Man Ro

doi:10.1007/978-3-319-73603-7_10

Convolution with Logarithmic Filter Groups for Efficient Shallow CNN

Tae Kwan Lee
, Wissam J. Baddar
, Seong Tae Kim
, Yong Man Ro

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

10 Citations (Scopus)

Abstract

In convolutional neural networks (CNNs), the filter grouping in convolution layers is known to be useful to reduce the network parameter size. In this paper, we propose a new logarithmic filter grouping which can capture the nonlinearity of filter distribution in CNNs. The proposed logarithmic filter grouping is installed in shallow CNNs applicable in a mobile application. Experiments were performed with the shallow CNNs for classification tasks. Our classification results on Multi-PIE dataset for facial expression recognition and CIFAR-10 dataset for object classification reveal that the compact CNN with the proposed logarithmic filter grouping scheme outperforms the same network with the uniform filter grouping in terms of accuracy and parameter efficiency. Our results indicate that the efficiency of shallow CNNs can be improved by the proposed logarithmic filter grouping.

Original language	English
Title of host publication	MultiMedia Modeling - 24th International Conference, MMM 2018, Proceedings
Editors	Klaus Schoeffmann, Moncef Gabbouj, Noel E. O'Connor, Ahmed Elgammal, Thanarat H. Chalidabhongse, Supavadee Aramvith, Chong Wah Ngo, Yo-Sung Ho
Publisher	Springer Verlag
Pages	117-129
Number of pages	13
ISBN (Print)	9783319736020
DOIs	https://doi.org/10.1007/978-3-319-73603-7_10
Publication status	Published - 2018
Event	24th International Conference on MultiMedia Modeling, MMM 2018 - Bangkok, Thailand Duration: 5 Feb 2018 → 7 Feb 2018

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	10704 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	24th International Conference on MultiMedia Modeling, MMM 2018
Country/Territory	Thailand
City	Bangkok
Period	5/02/18 → 7/02/18

Bibliographical note

Publisher Copyright:
© 2018, Springer International Publishing AG.

Keywords

CNN parameter efficiency
Classification tasks
Nonlinear logarithmic filter groups
Shallow convolutional neural network (CNN)

Access to Document

10.1007/978-3-319-73603-7_10

Cite this

Lee, T. K., Baddar, W. J., Kim, S. T., & Ro, Y. M. (2018). Convolution with Logarithmic Filter Groups for Efficient Shallow CNN. In K. Schoeffmann, M. Gabbouj, N. E. O'Connor, A. Elgammal, T. H. Chalidabhongse, S. Aramvith, C. W. Ngo, & Y.-S. Ho (Eds.), MultiMedia Modeling - 24th International Conference, MMM 2018, Proceedings (pp. 117-129). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10704 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-73603-7_10

Lee, Tae Kwan ; Baddar, Wissam J. ; Kim, Seong Tae et al. / Convolution with Logarithmic Filter Groups for Efficient Shallow CNN. MultiMedia Modeling - 24th International Conference, MMM 2018, Proceedings. editor / Klaus Schoeffmann ; Moncef Gabbouj ; Noel E. O'Connor ; Ahmed Elgammal ; Thanarat H. Chalidabhongse ; Supavadee Aramvith ; Chong Wah Ngo ; Yo-Sung Ho. Springer Verlag, 2018. pp. 117-129 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{f529792f46ed49aca0365a815ecb91d9,

title = "Convolution with Logarithmic Filter Groups for Efficient Shallow CNN",

abstract = "In convolutional neural networks (CNNs), the filter grouping in convolution layers is known to be useful to reduce the network parameter size. In this paper, we propose a new logarithmic filter grouping which can capture the nonlinearity of filter distribution in CNNs. The proposed logarithmic filter grouping is installed in shallow CNNs applicable in a mobile application. Experiments were performed with the shallow CNNs for classification tasks. Our classification results on Multi-PIE dataset for facial expression recognition and CIFAR-10 dataset for object classification reveal that the compact CNN with the proposed logarithmic filter grouping scheme outperforms the same network with the uniform filter grouping in terms of accuracy and parameter efficiency. Our results indicate that the efficiency of shallow CNNs can be improved by the proposed logarithmic filter grouping.",

keywords = "CNN parameter efficiency, Classification tasks, Nonlinear logarithmic filter groups, Shallow convolutional neural network (CNN)",

author = "Lee, \{Tae Kwan\} and Baddar, \{Wissam J.\} and Kim, \{Seong Tae\} and Ro, \{Yong Man\}",

note = "Publisher Copyright: {\textcopyright} 2018, Springer International Publishing AG.; 24th International Conference on MultiMedia Modeling, MMM 2018 ; Conference date: 05-02-2018 Through 07-02-2018",

year = "2018",

doi = "10.1007/978-3-319-73603-7\_10",

language = "English",

isbn = "9783319736020",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "117--129",

editor = "Klaus Schoeffmann and Moncef Gabbouj and O'Connor, \{Noel E.\} and Ahmed Elgammal and Chalidabhongse, \{Thanarat H.\} and Supavadee Aramvith and Ngo, \{Chong Wah\} and Yo-Sung Ho",

booktitle = "MultiMedia Modeling - 24th International Conference, MMM 2018, Proceedings",

address = "Germany",

}

Lee, TK, Baddar, WJ, Kim, ST & Ro, YM 2018, Convolution with Logarithmic Filter Groups for Efficient Shallow CNN. in K Schoeffmann, M Gabbouj, NE O'Connor, A Elgammal, TH Chalidabhongse, S Aramvith, CW Ngo & Y-S Ho (eds), MultiMedia Modeling - 24th International Conference, MMM 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10704 LNCS, Springer Verlag, pp. 117-129, 24th International Conference on MultiMedia Modeling, MMM 2018, Bangkok, Thailand, 5/02/18. https://doi.org/10.1007/978-3-319-73603-7_10

Convolution with Logarithmic Filter Groups for Efficient Shallow CNN. / Lee, Tae Kwan; Baddar, Wissam J.; Kim, Seong Tae et al.
MultiMedia Modeling - 24th International Conference, MMM 2018, Proceedings. ed. / Klaus Schoeffmann; Moncef Gabbouj; Noel E. O'Connor; Ahmed Elgammal; Thanarat H. Chalidabhongse; Supavadee Aramvith; Chong Wah Ngo; Yo-Sung Ho. Springer Verlag, 2018. p. 117-129 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10704 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Convolution with Logarithmic Filter Groups for Efficient Shallow CNN

AU - Lee, Tae Kwan

AU - Baddar, Wissam J.

AU - Kim, Seong Tae

AU - Ro, Yong Man

PY - 2018

Y1 - 2018

N2 - In convolutional neural networks (CNNs), the filter grouping in convolution layers is known to be useful to reduce the network parameter size. In this paper, we propose a new logarithmic filter grouping which can capture the nonlinearity of filter distribution in CNNs. The proposed logarithmic filter grouping is installed in shallow CNNs applicable in a mobile application. Experiments were performed with the shallow CNNs for classification tasks. Our classification results on Multi-PIE dataset for facial expression recognition and CIFAR-10 dataset for object classification reveal that the compact CNN with the proposed logarithmic filter grouping scheme outperforms the same network with the uniform filter grouping in terms of accuracy and parameter efficiency. Our results indicate that the efficiency of shallow CNNs can be improved by the proposed logarithmic filter grouping.

AB - In convolutional neural networks (CNNs), the filter grouping in convolution layers is known to be useful to reduce the network parameter size. In this paper, we propose a new logarithmic filter grouping which can capture the nonlinearity of filter distribution in CNNs. The proposed logarithmic filter grouping is installed in shallow CNNs applicable in a mobile application. Experiments were performed with the shallow CNNs for classification tasks. Our classification results on Multi-PIE dataset for facial expression recognition and CIFAR-10 dataset for object classification reveal that the compact CNN with the proposed logarithmic filter grouping scheme outperforms the same network with the uniform filter grouping in terms of accuracy and parameter efficiency. Our results indicate that the efficiency of shallow CNNs can be improved by the proposed logarithmic filter grouping.

KW - CNN parameter efficiency

KW - Classification tasks

KW - Nonlinear logarithmic filter groups

KW - Shallow convolutional neural network (CNN)

UR - https://www.scopus.com/pages/publications/85042098653

U2 - 10.1007/978-3-319-73603-7_10

DO - 10.1007/978-3-319-73603-7_10

M3 - Conference contribution

AN - SCOPUS:85042098653

SN - 9783319736020

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 117

EP - 129

BT - MultiMedia Modeling - 24th International Conference, MMM 2018, Proceedings

A2 - Schoeffmann, Klaus

A2 - Gabbouj, Moncef

A2 - O'Connor, Noel E.

A2 - Elgammal, Ahmed

A2 - Chalidabhongse, Thanarat H.

A2 - Aramvith, Supavadee

A2 - Ngo, Chong Wah

A2 - Ho, Yo-Sung

PB - Springer Verlag

T2 - 24th International Conference on MultiMedia Modeling, MMM 2018

Y2 - 5 February 2018 through 7 February 2018

ER -

Lee TK, Baddar WJ, Kim ST, Ro YM. Convolution with Logarithmic Filter Groups for Efficient Shallow CNN. In Schoeffmann K, Gabbouj M, O'Connor NE, Elgammal A, Chalidabhongse TH, Aramvith S, Ngo CW, Ho YS, editors, MultiMedia Modeling - 24th International Conference, MMM 2018, Proceedings. Springer Verlag. 2018. p. 117-129. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-73603-7_10

Convolution with Logarithmic Filter Groups for Efficient Shallow CNN

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this