Details
Original language | English |
---|---|
Title of host publication | Discovery Science |
Subtitle of host publication | 23rd International Conference, DS 2020, Proceedings |
Editors | Annalisa Appice, Grigorios Tsoumakas, Yannis Manolopoulos, Stan Matwin |
Publisher | Springer Science and Business Media Deutschland GmbH |
Pages | 159-174 |
Number of pages | 16 |
ISBN (electronic) | 9783030615277 |
ISBN (print) | 9783030615260 |
Publication status | Published - 15 Oct 2020 |
Event | 23rd International Conference on Discovery Science, DS 2020 - Thessaloniki, Greece Duration: 19 Oct 2020 → 21 Oct 2020 |
Publication series
Name | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
---|---|
ISSN (Print) | 0302-9743 |
ISSN (electronic) | 1611-3349 |
Abstract
Data-driven algorithms are employed in many applications, in which data become available in a sequential order, forcing the update of the model with new instances. In such dynamic environments, in which the underlying data distributions might evolve with time, fairness-aware learning cannot be considered as a one-off requirement, but rather it should comprise a continual requirement over the stream. Recent fairness-aware stream classifiers ignore the problem of class distribution skewness. As a result, such methods mitigate discrimination by “rejecting” minority instances at large due to their inability to effectively learn all classes. In this work, we propose, an online fairness-aware approach that maintains a valid and fair classifier over a stream is an online boosting approach that changes the training distribution in an online fashion based on both stream imbalance and discriminatory behavior of the model evaluated over the historical stream. Our experiments show that such long-term consideration of class-imbalance and fairness are beneficial for maintaining models that exhibit good predictive- and fairness-related performance.
Keywords
- Class-imbalance, Data streams, Fairness-aware classification
ASJC Scopus subject areas
- Mathematics(all)
- Theoretical Computer Science
- Computer Science(all)
- General Computer Science
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
Discovery Science : 23rd International Conference, DS 2020, Proceedings. ed. / Annalisa Appice; Grigorios Tsoumakas; Yannis Manolopoulos; Stan Matwin. Springer Science and Business Media Deutschland GmbH, 2020. p. 159-174 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
Research output: Chapter in book/report/conference proceeding › Conference contribution › Research › peer review
}
TY - GEN
T1 - Online Fairness-Aware Learning Under Class Imbalance
AU - Iosifidis, Vasileios
AU - Ntoutsi, Eirini
PY - 2020/10/15
Y1 - 2020/10/15
N2 - Data-driven algorithms are employed in many applications, in which data become available in a sequential order, forcing the update of the model with new instances. In such dynamic environments, in which the underlying data distributions might evolve with time, fairness-aware learning cannot be considered as a one-off requirement, but rather it should comprise a continual requirement over the stream. Recent fairness-aware stream classifiers ignore the problem of class distribution skewness. As a result, such methods mitigate discrimination by “rejecting” minority instances at large due to their inability to effectively learn all classes. In this work, we propose, an online fairness-aware approach that maintains a valid and fair classifier over a stream is an online boosting approach that changes the training distribution in an online fashion based on both stream imbalance and discriminatory behavior of the model evaluated over the historical stream. Our experiments show that such long-term consideration of class-imbalance and fairness are beneficial for maintaining models that exhibit good predictive- and fairness-related performance.
AB - Data-driven algorithms are employed in many applications, in which data become available in a sequential order, forcing the update of the model with new instances. In such dynamic environments, in which the underlying data distributions might evolve with time, fairness-aware learning cannot be considered as a one-off requirement, but rather it should comprise a continual requirement over the stream. Recent fairness-aware stream classifiers ignore the problem of class distribution skewness. As a result, such methods mitigate discrimination by “rejecting” minority instances at large due to their inability to effectively learn all classes. In this work, we propose, an online fairness-aware approach that maintains a valid and fair classifier over a stream is an online boosting approach that changes the training distribution in an online fashion based on both stream imbalance and discriminatory behavior of the model evaluated over the historical stream. Our experiments show that such long-term consideration of class-imbalance and fairness are beneficial for maintaining models that exhibit good predictive- and fairness-related performance.
KW - Class-imbalance
KW - Data streams
KW - Fairness-aware classification
UR - http://www.scopus.com/inward/record.url?scp=85094157142&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-61527-7_11
DO - 10.1007/978-3-030-61527-7_11
M3 - Conference contribution
AN - SCOPUS:85094157142
SN - 9783030615260
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 159
EP - 174
BT - Discovery Science
A2 - Appice, Annalisa
A2 - Tsoumakas, Grigorios
A2 - Manolopoulos, Yannis
A2 - Matwin, Stan
PB - Springer Science and Business Media Deutschland GmbH
T2 - 23rd International Conference on Discovery Science, DS 2020
Y2 - 19 October 2020 through 21 October 2020
ER -