Restoring the Generalizability of SVM Based Decoding in High Dimensional Neuroimage Data

Trine Julie Abrahamsen; Lars Kai Hansen

doi:10.1007/978-3-642-34713-9_32

Restoring the Generalizability of SVM Based Decoding in High Dimensional Neuroimage Data

Trine Julie Abrahamsen, Lars Kai Hansen

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

Abstract

Variance inflation is caused by a mismatch between linear projections of test and training data when projections are estimated on training sets smaller than the dimensionality of the feature space. We demonstrate that variance inflation can lead to an increased neuroimage decoding error rate for Support Vector Machines. However, good generalization may be recovered in part by a simple renormalization procedure. We show that with proper renormalization, cross-validation based parameter optimization leads to the acceptance of more non-linearity in neuroimage classifiers than would have been obtained without renormalization.

Original language	English
Title of host publication	Machine Learning and Interpretation in Neuroimaging : International Workshop, MLINI 2011, Held at NIPS 2011, Sierra Nevada, Spain, December 16-17, 2011, Revised Selected and Invited Contributions
Publisher	Springer
Publication date	2012
Pages	256-263
ISBN (Print)	978-3-642-34712-2
ISBN (Electronic)	978-3-642-34713-9
DOIs	https://doi.org/10.1007/978-3-642-34713-9_32
Publication status	Published - 2012
Event	International Workshop on Machine Learning and Interpretation in Neuroimaging (MLINI 2011) - Granada, Spain Duration: 16 Dec 2011 → 17 Dec 2011 https://sites.google.com/site/mlini2011/home

Workshop

Workshop	International Workshop on Machine Learning and Interpretation in Neuroimaging (MLINI 2011)
Country/Territory	Spain
City	Granada
Period	16/12/2011 → 17/12/2011
Internet address	https://sites.google.com/site/mlini2011/home

Series	Lecture Notes in Computer Science
Volume	7263
ISSN	0302-9743

Keywords

Support Vector Machines
Generalizability
Variance inflation
Imbalanced data

Access to Document

10.1007/978-3-642-34713-9_32

OpenUrl availability

Full text

Cite this

Abrahamsen, T. J., & Hansen, L. K. (2012). Restoring the Generalizability of SVM Based Decoding in High Dimensional Neuroimage Data. In Machine Learning and Interpretation in Neuroimaging: International Workshop, MLINI 2011, Held at NIPS 2011, Sierra Nevada, Spain, December 16-17, 2011, Revised Selected and Invited Contributions (pp. 256-263). Springer. https://doi.org/10.1007/978-3-642-34713-9_32

Abrahamsen, Trine Julie ; Hansen, Lars Kai. / Restoring the Generalizability of SVM Based Decoding in High Dimensional Neuroimage Data. Machine Learning and Interpretation in Neuroimaging: International Workshop, MLINI 2011, Held at NIPS 2011, Sierra Nevada, Spain, December 16-17, 2011, Revised Selected and Invited Contributions. Springer, 2012. pp. 256-263 (Lecture Notes in Computer Science, Vol. 7263).

@inproceedings{71c45737727e4b4b9c53101ed809a2cb,

title = "Restoring the Generalizability of SVM Based Decoding in High Dimensional Neuroimage Data",

abstract = "Variance inflation is caused by a mismatch between linear projections of test and training data when projections are estimated on training sets smaller than the dimensionality of the feature space. We demonstrate that variance inflation can lead to an increased neuroimage decoding error rate for Support Vector Machines. However, good generalization may be recovered in part by a simple renormalization procedure. We show that with proper renormalization, cross-validation based parameter optimization leads to the acceptance of more non-linearity in neuroimage classifiers than would have been obtained without renormalization.",

keywords = "Support Vector Machines, Generalizability, Variance inflation, Imbalanced data",

author = "Abrahamsen, {Trine Julie} and Hansen, {Lars Kai}",

year = "2012",

doi = "10.1007/978-3-642-34713-9_32",

language = "English",

isbn = "978-3-642-34712-2",

series = "Lecture Notes in Computer Science",

publisher = "Springer",

pages = "256--263",

booktitle = "Machine Learning and Interpretation in Neuroimaging",

note = "International Workshop on Machine Learning and Interpretation in Neuroimaging (MLINI 2011) ; Conference date: 16-12-2011 Through 17-12-2011",

url = "https://sites.google.com/site/mlini2011/home",

}

Abrahamsen, TJ & Hansen, LK 2012, Restoring the Generalizability of SVM Based Decoding in High Dimensional Neuroimage Data. in Machine Learning and Interpretation in Neuroimaging: International Workshop, MLINI 2011, Held at NIPS 2011, Sierra Nevada, Spain, December 16-17, 2011, Revised Selected and Invited Contributions. Springer, Lecture Notes in Computer Science, vol. 7263, pp. 256-263, International Workshop on Machine Learning and Interpretation in Neuroimaging (MLINI 2011), Granada, Spain, 16/12/2011. https://doi.org/10.1007/978-3-642-34713-9_32

Restoring the Generalizability of SVM Based Decoding in High Dimensional Neuroimage Data. / Abrahamsen, Trine Julie; Hansen, Lars Kai.
Machine Learning and Interpretation in Neuroimaging: International Workshop, MLINI 2011, Held at NIPS 2011, Sierra Nevada, Spain, December 16-17, 2011, Revised Selected and Invited Contributions. Springer, 2012. p. 256-263 (Lecture Notes in Computer Science, Vol. 7263).

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

TY - GEN

T1 - Restoring the Generalizability of SVM Based Decoding in High Dimensional Neuroimage Data

AU - Abrahamsen, Trine Julie

AU - Hansen, Lars Kai

PY - 2012

Y1 - 2012

N2 - Variance inflation is caused by a mismatch between linear projections of test and training data when projections are estimated on training sets smaller than the dimensionality of the feature space. We demonstrate that variance inflation can lead to an increased neuroimage decoding error rate for Support Vector Machines. However, good generalization may be recovered in part by a simple renormalization procedure. We show that with proper renormalization, cross-validation based parameter optimization leads to the acceptance of more non-linearity in neuroimage classifiers than would have been obtained without renormalization.

AB - Variance inflation is caused by a mismatch between linear projections of test and training data when projections are estimated on training sets smaller than the dimensionality of the feature space. We demonstrate that variance inflation can lead to an increased neuroimage decoding error rate for Support Vector Machines. However, good generalization may be recovered in part by a simple renormalization procedure. We show that with proper renormalization, cross-validation based parameter optimization leads to the acceptance of more non-linearity in neuroimage classifiers than would have been obtained without renormalization.

KW - Support Vector Machines

KW - Generalizability

KW - Variance inflation

KW - Imbalanced data

U2 - 10.1007/978-3-642-34713-9_32

DO - 10.1007/978-3-642-34713-9_32

M3 - Article in proceedings

SN - 978-3-642-34712-2

T3 - Lecture Notes in Computer Science

SP - 256

EP - 263

BT - Machine Learning and Interpretation in Neuroimaging

PB - Springer

T2 - International Workshop on Machine Learning and Interpretation in Neuroimaging (MLINI 2011)

Y2 - 16 December 2011 through 17 December 2011

ER -

Abrahamsen TJ, Hansen LK. Restoring the Generalizability of SVM Based Decoding in High Dimensional Neuroimage Data. In Machine Learning and Interpretation in Neuroimaging: International Workshop, MLINI 2011, Held at NIPS 2011, Sierra Nevada, Spain, December 16-17, 2011, Revised Selected and Invited Contributions. Springer. 2012. p. 256-263. (Lecture Notes in Computer Science, Vol. 7263). doi: 10.1007/978-3-642-34713-9_32

Restoring the Generalizability of SVM Based Decoding in High Dimensional Neuroimage Data

Abstract

Workshop

Keywords

Access to Document

OpenUrl availability

Fingerprint

Cite this