Weight Sharing and Deep Learning for Spectral Data

Jacob Sogaard Larsen; Line Clemmensen

doi:10.1109/ICASSP40776.2020.9053918

Weight Sharing and Deep Learning for Spectral Data

Jacob Sogaard Larsen
, Line Clemmensen

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

Abstract

We propose a novel method to co-train deep convolutional neural networks for data sets of differing position specific data. This is an advantage in chemometrics where individual measurements represent exact chemical compounds, e.g. for given wavelengths, and thus signals cannot be translated or resized without disturbing their interpretation. Our approach outperforms transfer learning for three small data sets co-trained with a medium sized data set.

Original language	English
Title of host publication	Proceedings of 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing
Publisher	IEEE
Publication date	May 2020
Pages	4227-4231
Article number	9053918
ISBN (Electronic)	9781509066315
DOIs	https://doi.org/10.1109/ICASSP40776.2020.9053918
Publication status	Published - May 2020
Event	2020 IEEE International Conference on Acoustics, Speech, and Signal Processing - Virtual event, Barcelona, Spain Duration: 4 May 2020 → 8 May 2020 Conference number: 45

Conference

Conference	2020 IEEE International Conference on Acoustics, Speech, and Signal Processing
Number	45
Location	Virtual event
Country/Territory	Spain
City	Barcelona
Period	04/05/2020 → 08/05/2020
Sponsor	IEEE

Series	ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Volume	2020-May
ISSN	1520-6149

Keywords

Co-training
Deep Learning
Spectroscopic data
Transfer Learning
Weight Sharing

Access to Document

10.1109/ICASSP40776.2020.9053918

OpenUrl availability

Check availability in DTU Findit

Cite this

@inproceedings{ebce48f0dc484030b0360eaf5aa4ea96,

title = "Weight Sharing and Deep Learning for Spectral Data",

abstract = "We propose a novel method to co-train deep convolutional neural networks for data sets of differing position specific data. This is an advantage in chemometrics where individual measurements represent exact chemical compounds, e.g. for given wavelengths, and thus signals cannot be translated or resized without disturbing their interpretation. Our approach outperforms transfer learning for three small data sets co-trained with a medium sized data set.",

keywords = "Co-training, Deep Learning, Spectroscopic data, Transfer Learning, Weight Sharing",

author = "Larsen, \{Jacob Sogaard\} and Line Clemmensen",

year = "2020",

month = may,

doi = "10.1109/ICASSP40776.2020.9053918",

language = "English",

series = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",

publisher = "IEEE",

pages = "4227--4231",

booktitle = "Proceedings of 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing",

address = "United States",

note = "2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2020 ; Conference date: 04-05-2020 Through 08-05-2020",

}

Larsen, JS & Clemmensen, L 2020, Weight Sharing and Deep Learning for Spectral Data. in Proceedings of 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing., 9053918, IEEE, ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings, vol. 2020-May, pp. 4227-4231, 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, Barcelona, Spain, 04/05/2020. https://doi.org/10.1109/ICASSP40776.2020.9053918

Weight Sharing and Deep Learning for Spectral Data. / Larsen, Jacob Sogaard; Clemmensen, Line.
Proceedings of 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing. IEEE, 2020. p. 4227-4231 9053918 (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings, Vol. 2020-May).

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

TY - GEN

T1 - Weight Sharing and Deep Learning for Spectral Data

AU - Larsen, Jacob Sogaard

AU - Clemmensen, Line

N1 - Conference code: 45

PY - 2020/5

Y1 - 2020/5

N2 - We propose a novel method to co-train deep convolutional neural networks for data sets of differing position specific data. This is an advantage in chemometrics where individual measurements represent exact chemical compounds, e.g. for given wavelengths, and thus signals cannot be translated or resized without disturbing their interpretation. Our approach outperforms transfer learning for three small data sets co-trained with a medium sized data set.

AB - We propose a novel method to co-train deep convolutional neural networks for data sets of differing position specific data. This is an advantage in chemometrics where individual measurements represent exact chemical compounds, e.g. for given wavelengths, and thus signals cannot be translated or resized without disturbing their interpretation. Our approach outperforms transfer learning for three small data sets co-trained with a medium sized data set.

KW - Co-training

KW - Deep Learning

KW - Spectroscopic data

KW - Transfer Learning

KW - Weight Sharing

U2 - 10.1109/ICASSP40776.2020.9053918

DO - 10.1109/ICASSP40776.2020.9053918

M3 - Article in proceedings

AN - SCOPUS:85089224015

T3 - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

SP - 4227

EP - 4231

BT - Proceedings of 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing

PB - IEEE

T2 - 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing

Y2 - 4 May 2020 through 8 May 2020

ER -

Weight Sharing and Deep Learning for Spectral Data

Abstract

Conference

Keywords

Access to Document

OpenUrl availability

Fingerprint

Cite this