Target robust discriminant analysis

Wouter M. Kouw; Marco Loog

doi:https://doi.org/10.1007/978-3-030-73973-7_1

Target robust discriminant analysis

Wouter M. Kouw, Marco Loog

Signal Processing Systems

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

Abstract

In practice, the data distribution at test time often differs, to a smaller or larger extent, from that of the original training data. Consequentially, the so-called source classifier, trained on the available labelled data, deteriorates on the test, or target, data. Domain adaptive classifiers aim to combat this problem, but typically assume some particular form of domain shift. Most are not robust to violations of domain shift assumptions and may even perform worse than their non-adaptive counterparts. We construct robust parameter estimators for discriminant analysis that guarantee performance improvements of the adaptive classifier over the non-adaptive source classifier.

Original language	English
Title of host publication	Structural, Syntactic, and Statistical Pattern Recognition
Subtitle of host publication	Joint IAPR International Workshops, S+SSPR 2020, Proceedings
Editors	Andrea Torsello, Luca Rossi, Marcello Pelillo, Battista Biggio, Antonio Robles-Kelly
Place of Publication	Cham
Publisher	Springer Nature
Chapter	1
Pages	3-13
Number of pages	11
ISBN (Electronic)	978-3-030-73973-7
ISBN (Print)	978-3-030-73972-0
DOIs	https://doi.org/10.1007/978-3-030-73973-7_1
Publication status	Published - 10 Apr 2021
Event	Joint IAPR International Workshops, S+SSPR 2020, Padua, Italy, January 21–22, 2021 - Padua, Italy Duration: 21 Jan 2021 → 22 Jan 2021

Publication series

Name	Lecture Notes in Computer Science (LNCS)
Volume	12644
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Name	Image Processing, Computer Vision, Pattern Recognition, and Graphics (LNIP)
Volume	12644

Conference

Conference	Joint IAPR International Workshops, S+SSPR 2020, Padua, Italy, January 21–22, 2021
Country/Territory	Italy
City	Padua
Period	21/01/21 → 22/01/21

Keywords

Domain adaptation
Robust estimators
Discriminant analysis
Robustness

Access to Document

https://doi.org/10.1007/978-3-030-73973-7_1

https://arxiv.org/abs/1806.09463

Cite this

Kouw, W. M., & Loog, M. (2021). Target robust discriminant analysis. In A. Torsello, L. Rossi, M. Pelillo, B. Biggio, & A. Robles-Kelly (Eds.), Structural, Syntactic, and Statistical Pattern Recognition : Joint IAPR International Workshops, S+SSPR 2020, Proceedings (pp. 3-13). (Lecture Notes in Computer Science (LNCS); Vol. 12644), ( Image Processing, Computer Vision, Pattern Recognition, and Graphics (LNIP); Vol. 12644). Springer Nature. https://doi.org/10.1007/978-3-030-73973-7_1

Kouw, Wouter M. ; Loog, Marco. / Target robust discriminant analysis. Structural, Syntactic, and Statistical Pattern Recognition : Joint IAPR International Workshops, S+SSPR 2020, Proceedings. editor / Andrea Torsello ; Luca Rossi ; Marcello Pelillo ; Battista Biggio ; Antonio Robles-Kelly. Cham : Springer Nature, 2021. pp. 3-13 (Lecture Notes in Computer Science (LNCS)). ( Image Processing, Computer Vision, Pattern Recognition, and Graphics (LNIP)).

@inproceedings{1feb0b37024740dfa82e1d18a36ee634,

title = "Target robust discriminant analysis",

abstract = "In practice, the data distribution at test time often differs, to a smaller or larger extent, from that of the original training data. Consequentially, the so-called source classifier, trained on the available labelled data, deteriorates on the test, or target, data. Domain adaptive classifiers aim to combat this problem, but typically assume some particular form of domain shift. Most are not robust to violations of domain shift assumptions and may even perform worse than their non-adaptive counterparts. We construct robust parameter estimators for discriminant analysis that guarantee performance improvements of the adaptive classifier over the non-adaptive source classifier.",

keywords = "Domain adaptation, Robust estimators, Discriminant analysis, Robustness",

author = "Kouw, {Wouter M.} and Marco Loog",

year = "2021",

month = apr,

day = "10",

doi = "https://doi.org/10.1007/978-3-030-73973-7_1",

language = "English",

isbn = "978-3-030-73972-0",

series = "Lecture Notes in Computer Science (LNCS)",

publisher = "Springer Nature",

pages = "3--13",

editor = "Andrea Torsello and Luca Rossi and Marcello Pelillo and Battista Biggio and Antonio Robles-Kelly",

booktitle = "Structural, Syntactic, and Statistical Pattern Recognition",

address = "Singapore",

note = "Joint IAPR International Workshops, S+SSPR 2020, Padua, Italy, January 21–22, 2021 ; Conference date: 21-01-2021 Through 22-01-2021",

}

Kouw, WM & Loog, M 2021, Target robust discriminant analysis. in A Torsello, L Rossi, M Pelillo, B Biggio & A Robles-Kelly (eds), Structural, Syntactic, and Statistical Pattern Recognition : Joint IAPR International Workshops, S+SSPR 2020, Proceedings. Lecture Notes in Computer Science (LNCS), vol. 12644, Image Processing, Computer Vision, Pattern Recognition, and Graphics (LNIP), vol. 12644, Springer Nature, Cham, pp. 3-13, Joint IAPR International Workshops, S+SSPR 2020, Padua, Italy, January 21–22, 2021, Padua, Italy, 21/01/21. https://doi.org/10.1007/978-3-030-73973-7_1

Target robust discriminant analysis. / Kouw, Wouter M.; Loog, Marco.

Structural, Syntactic, and Statistical Pattern Recognition : Joint IAPR International Workshops, S+SSPR 2020, Proceedings. ed. / Andrea Torsello; Luca Rossi; Marcello Pelillo; Battista Biggio; Antonio Robles-Kelly. Cham : Springer Nature, 2021. p. 3-13 (Lecture Notes in Computer Science (LNCS); Vol. 12644), ( Image Processing, Computer Vision, Pattern Recognition, and Graphics (LNIP); Vol. 12644).

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

TY - GEN

T1 - Target robust discriminant analysis

AU - Kouw, Wouter M.

AU - Loog, Marco

PY - 2021/4/10

Y1 - 2021/4/10

N2 - In practice, the data distribution at test time often differs, to a smaller or larger extent, from that of the original training data. Consequentially, the so-called source classifier, trained on the available labelled data, deteriorates on the test, or target, data. Domain adaptive classifiers aim to combat this problem, but typically assume some particular form of domain shift. Most are not robust to violations of domain shift assumptions and may even perform worse than their non-adaptive counterparts. We construct robust parameter estimators for discriminant analysis that guarantee performance improvements of the adaptive classifier over the non-adaptive source classifier.

AB - In practice, the data distribution at test time often differs, to a smaller or larger extent, from that of the original training data. Consequentially, the so-called source classifier, trained on the available labelled data, deteriorates on the test, or target, data. Domain adaptive classifiers aim to combat this problem, but typically assume some particular form of domain shift. Most are not robust to violations of domain shift assumptions and may even perform worse than their non-adaptive counterparts. We construct robust parameter estimators for discriminant analysis that guarantee performance improvements of the adaptive classifier over the non-adaptive source classifier.

KW - Domain adaptation

KW - Robust estimators

KW - Discriminant analysis

KW - Robustness

UR - http://www.scopus.com/inward/record.url?scp=85104734662&partnerID=8YFLogxK

U2 - https://doi.org/10.1007/978-3-030-73973-7_1

DO - https://doi.org/10.1007/978-3-030-73973-7_1

M3 - Conference contribution

SN - 978-3-030-73972-0

T3 - Lecture Notes in Computer Science (LNCS)

SP - 3

EP - 13

BT - Structural, Syntactic, and Statistical Pattern Recognition

A2 - Torsello, Andrea

A2 - Rossi, Luca

A2 - Pelillo, Marcello

A2 - Biggio, Battista

A2 - Robles-Kelly, Antonio

PB - Springer Nature

CY - Cham

T2 - Joint IAPR International Workshops, S+SSPR 2020, Padua, Italy, January 21–22, 2021

Y2 - 21 January 2021 through 22 January 2021

ER -

Kouw WM, Loog M. Target robust discriminant analysis. In Torsello A, Rossi L, Pelillo M, Biggio B, Robles-Kelly A, editors, Structural, Syntactic, and Statistical Pattern Recognition : Joint IAPR International Workshops, S+SSPR 2020, Proceedings. Cham: Springer Nature. 2021. p. 3-13. (Lecture Notes in Computer Science (LNCS)). ( Image Processing, Computer Vision, Pattern Recognition, and Graphics (LNIP)). https://doi.org/10.1007/978-3-030-73973-7_1

Target robust discriminant analysis

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this