Data driven constraints for the SVM

Sune Darkner; Line H. Clemmensen

doi:10.1007/978-3-642-35428-1_9

Data driven constraints for the SVM

Sune Darkner, Line H. Clemmensen

Department of Computer Science

Abstract

We propose a generalized data driven constraint for support vector machines exemplified by classification of paired observations in general and specifically on the human ear canal. This is particularly interesting in dynamic cases such as tissue movement or pathologies developing over time. Assuming that two observations of the same subject in different states span a vector, we hypothesise that such structure of the data contains implicit information which can aid the classification, thus the name data driven constraints. We derive a constraint based on the data which allow for the use of the ℓ₁- norm on the constraint while still allowing for the application of kernels. We specialize the proposed constraint to orthogonality of the vectors between paired observations and the estimated hyperplane. We show that imposing the constraint of orthogonality on the paired data yields a more robust classifier solution, compared to the SVM i.e. reduces variance and improves classification rates. We present a quantitative measure of the information level contained in the pairing and test the method on simulated as well as a high-dimensional paired data set of ear-canal surfaces.

Original language	English
Title of host publication	Machine Learning in Medical Imaging : Third International Workshop, MLMI 2012, Held in Conjunction with MICCAI 2012, Nice, France, October 1, 2012, Revised Selected Papers
Editors	Fei Wang, Dinggang Shen, Pingkun Yan, Kenji Suzuki
Number of pages	8
Publisher	Springer
Publication date	2012
Pages	70-77
ISBN (Print)	978-3-642-35427-4
ISBN (Electronic)	978-3-642-35428-1
DOIs	https://doi.org/10.1007/978-3-642-35428-1_9
Publication status	Published - 2012
Event	3rd International Workshop on Machine Learning in Medical Imaging - Nice, France Duration: 1 Oct 2012 → 1 Oct 2012 Conference number: 3

Conference

Conference	3rd International Workshop on Machine Learning in Medical Imaging
Number	3
Country/Territory	France
City	Nice
Period	01/10/2012 → 01/10/2012

Series	Lecture notes in computer science
Volume	7588
ISSN	0302-9743

Access to Document

10.1007/978-3-642-35428-1_9

Cite this

Darkner, S., & Clemmensen, L. H. (2012). Data driven constraints for the SVM. In F. Wang, D. Shen, P. Yan, & K. Suzuki (Eds.), Machine Learning in Medical Imaging: Third International Workshop, MLMI 2012, Held in Conjunction with MICCAI 2012, Nice, France, October 1, 2012, Revised Selected Papers (pp. 70-77). Springer. Lecture notes in computer science Vol. 7588 https://doi.org/10.1007/978-3-642-35428-1_9

Data driven constraints for the SVM. / Darkner, Sune; Clemmensen, Line H.

Machine Learning in Medical Imaging: Third International Workshop, MLMI 2012, Held in Conjunction with MICCAI 2012, Nice, France, October 1, 2012, Revised Selected Papers. ed. / Fei Wang; Dinggang Shen; Pingkun Yan; Kenji Suzuki. Springer, 2012. p. 70-77 (Lecture notes in computer science, Vol. 7588).

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

Darkner, S & Clemmensen, LH 2012, Data driven constraints for the SVM. in F Wang, D Shen, P Yan & K Suzuki (eds), Machine Learning in Medical Imaging: Third International Workshop, MLMI 2012, Held in Conjunction with MICCAI 2012, Nice, France, October 1, 2012, Revised Selected Papers. Springer, Lecture notes in computer science, vol. 7588, pp. 70-77, 3rd International Workshop on Machine Learning in Medical Imaging, Nice, France, 01/10/2012. https://doi.org/10.1007/978-3-642-35428-1_9

Darkner S, Clemmensen LH. Data driven constraints for the SVM. In Wang F, Shen D, Yan P, Suzuki K, editors, Machine Learning in Medical Imaging: Third International Workshop, MLMI 2012, Held in Conjunction with MICCAI 2012, Nice, France, October 1, 2012, Revised Selected Papers. Springer. 2012. p. 70-77. (Lecture notes in computer science, Vol. 7588). doi: 10.1007/978-3-642-35428-1_9

Darkner, Sune ; Clemmensen, Line H. / Data driven constraints for the SVM. Machine Learning in Medical Imaging: Third International Workshop, MLMI 2012, Held in Conjunction with MICCAI 2012, Nice, France, October 1, 2012, Revised Selected Papers. editor / Fei Wang ; Dinggang Shen ; Pingkun Yan ; Kenji Suzuki. Springer, 2012. pp. 70-77 (Lecture notes in computer science, Vol. 7588).

@inproceedings{7a61dadbec764030b37ea84e5b0b79bd,

title = "Data driven constraints for the SVM",

abstract = "We propose a generalized data driven constraint for support vector machines exemplified by classification of paired observations in general and specifically on the human ear canal. This is particularly interesting in dynamic cases such as tissue movement or pathologies developing over time. Assuming that two observations of the same subject in different states span a vector, we hypothesise that such structure of the data contains implicit information which can aid the classification, thus the name data driven constraints. We derive a constraint based on the data which allow for the use of the ℓ1- norm on the constraint while still allowing for the application of kernels. We specialize the proposed constraint to orthogonality of the vectors between paired observations and the estimated hyperplane. We show that imposing the constraint of orthogonality on the paired data yields a more robust classifier solution, compared to the SVM i.e. reduces variance and improves classification rates. We present a quantitative measure of the information level contained in the pairing and test the method on simulated as well as a high-dimensional paired data set of ear-canal surfaces.",

author = "Sune Darkner and Clemmensen, {Line H.}",

year = "2012",

doi = "10.1007/978-3-642-35428-1_9",

language = "English",

isbn = "978-3-642-35427-4",

series = "Lecture notes in computer science",

publisher = "Springer",

pages = "70--77",

editor = "Fei Wang and Dinggang Shen and Pingkun Yan and Kenji Suzuki",

booktitle = "Machine Learning in Medical Imaging",

note = "3rd International Workshop on Machine Learning in Medical Imaging, MLMI 2012 ; Conference date: 01-10-2012 Through 01-10-2012",

}

TY - GEN

T1 - Data driven constraints for the SVM

AU - Darkner, Sune

AU - Clemmensen, Line H.

N1 - Conference code: 3

PY - 2012

Y1 - 2012

N2 - We propose a generalized data driven constraint for support vector machines exemplified by classification of paired observations in general and specifically on the human ear canal. This is particularly interesting in dynamic cases such as tissue movement or pathologies developing over time. Assuming that two observations of the same subject in different states span a vector, we hypothesise that such structure of the data contains implicit information which can aid the classification, thus the name data driven constraints. We derive a constraint based on the data which allow for the use of the ℓ1- norm on the constraint while still allowing for the application of kernels. We specialize the proposed constraint to orthogonality of the vectors between paired observations and the estimated hyperplane. We show that imposing the constraint of orthogonality on the paired data yields a more robust classifier solution, compared to the SVM i.e. reduces variance and improves classification rates. We present a quantitative measure of the information level contained in the pairing and test the method on simulated as well as a high-dimensional paired data set of ear-canal surfaces.

AB - We propose a generalized data driven constraint for support vector machines exemplified by classification of paired observations in general and specifically on the human ear canal. This is particularly interesting in dynamic cases such as tissue movement or pathologies developing over time. Assuming that two observations of the same subject in different states span a vector, we hypothesise that such structure of the data contains implicit information which can aid the classification, thus the name data driven constraints. We derive a constraint based on the data which allow for the use of the ℓ1- norm on the constraint while still allowing for the application of kernels. We specialize the proposed constraint to orthogonality of the vectors between paired observations and the estimated hyperplane. We show that imposing the constraint of orthogonality on the paired data yields a more robust classifier solution, compared to the SVM i.e. reduces variance and improves classification rates. We present a quantitative measure of the information level contained in the pairing and test the method on simulated as well as a high-dimensional paired data set of ear-canal surfaces.

U2 - 10.1007/978-3-642-35428-1_9

DO - 10.1007/978-3-642-35428-1_9

M3 - Article in proceedings

SN - 978-3-642-35427-4

T3 - Lecture notes in computer science

SP - 70

EP - 77

BT - Machine Learning in Medical Imaging

A2 - Wang, Fei

A2 - Shen, Dinggang

A2 - Yan, Pingkun

A2 - Suzuki, Kenji

PB - Springer

T2 - 3rd International Workshop on Machine Learning in Medical Imaging

Y2 - 1 October 2012 through 1 October 2012

ER -

Data driven constraints for the SVM

Abstract

Conference

Access to Document

Fingerprint

Cite this