Teach to hash: A deep supervised hashing framework with data selection

Xiang Li; Chao Ma; Jie Yang; Yu Qiao

doi:10.1007/978-3-030-04167-0_11

Teach to hash: A deep supervised hashing framework with data selection

Xiang Li
, Chao Ma
, Jie Yang
, Yu Qiao

Shanghai Jiao Tong University

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

1 Scopus citations

Abstract

Recent years have witnessed wide applications of deep learning for large-scale image hashing tasks, as deep hashing algorithms can simultaneously learn feature representations and hash codes in an end-to-end way. However, although these methods have obtained promising results to some extent, they seldom take the effect of different training samples into account and treat all samples equally throughout the training procedure. Therefore, in this paper, we propose a novel deep hashing algorithm dubbed “Teach to Hash” (T2H), which introduces a “teacher” to automatically select the most effective samples for the current training period. To be specific, the “teacher” utilizes two criteria to measure the effectivity of all samples, and iteratively update the training set with the most effective ones. Experimental results on two typical image datasets indicate that the introduced “teacher” can significantly improve the performance of deep hashing framework and the proposed method outperforms the state-of-the-art hashing methods.

Original language	English
Title of host publication	Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings
Editors	Long Cheng, Andrew Chi Sing Leung, Seiichi Ozawa
Publisher	Springer Verlag
Pages	120-129
Number of pages	10
ISBN (Print)	9783030041663
DOIs	https://doi.org/10.1007/978-3-030-04167-0_11
State	Published - 2018
Externally published	Yes
Event	25th International Conference on Neural Information Processing, ICONIP 2018 - Siem Reap, Cambodia Duration: 13 Dec 2018 → 16 Dec 2018

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11301 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	25th International Conference on Neural Information Processing, ICONIP 2018
Country/Territory	Cambodia
City	Siem Reap
Period	13/12/18 → 16/12/18

Keywords

Data selection
Deep learning
Supervised hashing

Access to Document

10.1007/978-3-030-04167-0_11

Cite this

Li, X., Ma, C., Yang, J., & Qiao, Y. (2018). Teach to hash: A deep supervised hashing framework with data selection. In L. Cheng, A. C. S. Leung, & S. Ozawa (Eds.), Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings (pp. 120-129). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11301 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-04167-0_11

Li, Xiang ; Ma, Chao ; Yang, Jie et al. / Teach to hash : A deep supervised hashing framework with data selection. Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings. editor / Long Cheng ; Andrew Chi Sing Leung ; Seiichi Ozawa. Springer Verlag, 2018. pp. 120-129 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{4035fb21f11b4fce8ef99add02b1ed29,

title = "Teach to hash: A deep supervised hashing framework with data selection",

abstract = "Recent years have witnessed wide applications of deep learning for large-scale image hashing tasks, as deep hashing algorithms can simultaneously learn feature representations and hash codes in an end-to-end way. However, although these methods have obtained promising results to some extent, they seldom take the effect of different training samples into account and treat all samples equally throughout the training procedure. Therefore, in this paper, we propose a novel deep hashing algorithm dubbed {\textquotedblleft}Teach to Hash{\textquotedblright} (T2H), which introduces a {\textquotedblleft}teacher{\textquotedblright} to automatically select the most effective samples for the current training period. To be specific, the {\textquotedblleft}teacher{\textquotedblright} utilizes two criteria to measure the effectivity of all samples, and iteratively update the training set with the most effective ones. Experimental results on two typical image datasets indicate that the introduced {\textquotedblleft}teacher{\textquotedblright} can significantly improve the performance of deep hashing framework and the proposed method outperforms the state-of-the-art hashing methods.",

keywords = "Data selection, Deep learning, Supervised hashing",

author = "Xiang Li and Chao Ma and Jie Yang and Yu Qiao",

note = "Publisher Copyright: {\textcopyright} 2018, Springer Nature Switzerland AG.; 25th International Conference on Neural Information Processing, ICONIP 2018 ; Conference date: 13-12-2018 Through 16-12-2018",

year = "2018",

doi = "10.1007/978-3-030-04167-0\_11",

language = "英语",

isbn = "9783030041663",

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

publisher = "Springer Verlag",

pages = "120--129",

editor = "Long Cheng and Leung, \{Andrew Chi Sing\} and Seiichi Ozawa",

booktitle = "Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings",

}

Li, X, Ma, C, Yang, J & Qiao, Y 2018, Teach to hash: A deep supervised hashing framework with data selection. in L Cheng, ACS Leung & S Ozawa (eds), Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11301 LNCS, Springer Verlag, pp. 120-129, 25th International Conference on Neural Information Processing, ICONIP 2018, Siem Reap, Cambodia, 13/12/18. https://doi.org/10.1007/978-3-030-04167-0_11

Teach to hash: A deep supervised hashing framework with data selection. / Li, Xiang; Ma, Chao; Yang, Jie et al.
Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings. ed. / Long Cheng; Andrew Chi Sing Leung; Seiichi Ozawa. Springer Verlag, 2018. p. 120-129 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11301 LNCS).

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

TY - GEN

T1 - Teach to hash

T2 - 25th International Conference on Neural Information Processing, ICONIP 2018

AU - Li, Xiang

AU - Ma, Chao

AU - Yang, Jie

AU - Qiao, Yu

PY - 2018

Y1 - 2018

N2 - Recent years have witnessed wide applications of deep learning for large-scale image hashing tasks, as deep hashing algorithms can simultaneously learn feature representations and hash codes in an end-to-end way. However, although these methods have obtained promising results to some extent, they seldom take the effect of different training samples into account and treat all samples equally throughout the training procedure. Therefore, in this paper, we propose a novel deep hashing algorithm dubbed “Teach to Hash” (T2H), which introduces a “teacher” to automatically select the most effective samples for the current training period. To be specific, the “teacher” utilizes two criteria to measure the effectivity of all samples, and iteratively update the training set with the most effective ones. Experimental results on two typical image datasets indicate that the introduced “teacher” can significantly improve the performance of deep hashing framework and the proposed method outperforms the state-of-the-art hashing methods.

AB - Recent years have witnessed wide applications of deep learning for large-scale image hashing tasks, as deep hashing algorithms can simultaneously learn feature representations and hash codes in an end-to-end way. However, although these methods have obtained promising results to some extent, they seldom take the effect of different training samples into account and treat all samples equally throughout the training procedure. Therefore, in this paper, we propose a novel deep hashing algorithm dubbed “Teach to Hash” (T2H), which introduces a “teacher” to automatically select the most effective samples for the current training period. To be specific, the “teacher” utilizes two criteria to measure the effectivity of all samples, and iteratively update the training set with the most effective ones. Experimental results on two typical image datasets indicate that the introduced “teacher” can significantly improve the performance of deep hashing framework and the proposed method outperforms the state-of-the-art hashing methods.

KW - Data selection

KW - Deep learning

KW - Supervised hashing

UR - https://www.scopus.com/pages/publications/85059054968

U2 - 10.1007/978-3-030-04167-0_11

DO - 10.1007/978-3-030-04167-0_11

M3 - 会议稿件

AN - SCOPUS:85059054968

SN - 9783030041663

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

SP - 120

EP - 129

BT - Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings

A2 - Cheng, Long

A2 - Leung, Andrew Chi Sing

A2 - Ozawa, Seiichi

PB - Springer Verlag

Y2 - 13 December 2018 through 16 December 2018

ER -

Li X, Ma C, Yang J, Qiao Y. Teach to hash: A deep supervised hashing framework with data selection. In Cheng L, Leung ACS, Ozawa S, editors, Neural Information Processing - 25th International Conference, ICONIP 2018, Proceedings. Springer Verlag. 2018. p. 120-129. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-04167-0_11

Teach to hash: A deep supervised hashing framework with data selection

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this