Predictive Fair Representation Learning with Variational Autoencoders

Yamada, Tatsuya; Konishi, Takuya; Kawahara, Yoshinobu

doi:https://doi.org/10.1007/s00354-026-00320-0

Predictive Fair Representation Learning with Variational Autoencoders

Yamada, Tatsuya; Konishi, Takuya; Kawahara, Yoshinobu

New Generation Computing, 2026, 44(16)

Number of Access:0（2026-05-25 17:51 Counts）

Identifier to cite or link to this item: https://hdl.handle.net/11094/104796

Link to primary information

File	Format	Terms of use	Size	Views	Date.Available	Description	information
NewGenerComput_44_16	pdf	None	1.52 MB		2026.05.22

Creative Commons : Attribution

Item Information

Output File Export EndNote Basic Export Mendeley

Title	Predictive Fair Representation Learning with Variational Autoencoders
Creator	CreatorName Yamada, Tatsuya
	NRID 1000020760169 1000020760169 researchmap permalink takuya_konishi takuya_konishi Scopus Author ID 55389615000 55389615000 CreatorName Konishi, Takuya
	NRID 1000000514796 1000000514796 ORCID 0000-0001-7789-4709 0000-0001-7789-4709 researchmap permalink yoshinobu.kawahara yoshinobu.kawahara Scopus Author ID 9744960500 9744960500 CreatorName Kawahara, Yoshinobu
Subject	Contrastive loss
	Fair representation learning
	Oversampling
	Variational autoencoders
Description	Yamada T., Konishi T., Kawahara Y.. Predictive Fair Representation Learning with Variational Autoencoders. New Generation Computing 44 (2026); https://doi.org/10.1007/s00354-026-00320-0.
Abstract	Fairness is a critical concern in machine learning, as biases in prediction outcomes with respect to sensitive attributes such as gender and race have raised ethical and societal issues in real-world applications. To address this, fair representation learning aims to extract essential information from data as representations that are useful but independent of sensitive attributes. However, existing methods often suffer from limited predictive performance when their learned representations are applied to downstream tasks. In this study, we propose predictive FairDisCo (PdFairDisCo), an extension of FairDisCo to learn fair representations with variational autoencoders. PdFairDisCo enhances predictive performance by incorporating contrastive losses into the objective function. In addition, we introduce two oversampling methods for PdFairDisCo to mitigate bias by balancing the proportion of the sensitive attribute in training data. We demonstrate the effectiveness of PdFairDisCo through experiments. The experiments also show that the oversampling methods can further improve the performance of fairness.
Publisher	Springer
Source Title	New Generation Computing
Volume (Issue)	44(16)
Date of Issued	2026-03-25
Language	English
Handle URL	https://hdl.handle.net/11094/104796
PISSN	02883635
eISSN	18827055
Relation.isIdenticalTo	DOI (Publisher Version) https://doi.org/10.1007/s00354-026-00320-0
Access Rights	open access
CopyRight	This article is licensed under a Creative Commons Attribution 4.0 International License.
oaire:version	Version of Record
Category	学術雑誌論文 Journal Article

Resource Type	journal article
Local Type	Journal Article
DCMI Type	Text
DCTERMS.bibliographicCitation	New Generation Computing. 2026, 44(16)
DC.title	Predictive Fair Representation Learning with Variational Autoencoders
DC.creator	Yamada, Tatsuya
	Konishi, Takuya
	Kawahara, Yoshinobu
DC.publisher	Springer
DC.language' scheme='DCTERMS.RFC1766	English
DC.type' scheme='DCTERMS.DCMIType	Text
DCTERMS.issued' scheme='DCTERMS.W3CDTF	2026-03-25
DC.identifier	https://doi.org/10.1007/s00354-026-00320-0
DC.subject	Contrastive loss
	Fair representation learning
	Oversampling
	Variational autoencoders
DC.description	Yamada T., Konishi T., Kawahara Y.. Predictive Fair Representation Learning with Variational Autoencoders. New Generation Computing 44 (2026); https://doi.org/10.1007/s00354-026-00320-0.
DCTERMS.abstract	Fairness is a critical concern in machine learning, as biases in prediction outcomes with respect to sensitive attributes such as gender and race have raised ethical and societal issues in real-world applications. To address this, fair representation learning aims to extract essential information from data as representations that are useful but independent of sensitive attributes. However, existing methods often suffer from limited predictive performance when their learned representations are applied to downstream tasks. In this study, we propose predictive FairDisCo (PdFairDisCo), an extension of FairDisCo to learn fair representations with variational autoencoders. PdFairDisCo enhances predictive performance by incorporating contrastive losses into the objective function. In addition, we introduce two oversampling methods for PdFairDisCo to mitigate bias by balancing the proportion of the sensitive attribute in training data. We demonstrate the effectiveness of PdFairDisCo through experiments. The experiments also show that the oversampling methods can further improve the performance of fairness.
DC.rights	This article is licensed under a Creative Commons Attribution 4.0 International License.
DC.format	104796
citation_title	Predictive Fair Representation Learning with Variational Autoencoders
citation_author	Yamada, Tatsuya
	Konishi, Takuya
	Kawahara, Yoshinobu
citation_publisher	Springer
citation_language	English
citation_date	2026-03-25
citation_journal_title	New Generation Computing
citation_volume	44
citation_issue	16
citation_doi	https://doi.org/10.1007/s00354-026-00320-0
citation_public_url	https://hdl.handle.net/11094/104796

Predictive Fair Representation Learning with Variational Autoencoders

Link to primary information

Item Information

Show details