FairGAN: Fairness-aware Generative Adversarial Networks

Xu, Danfeng; Yuan, Shuhan; Zhang, Lu; Wu, Xintao

doi:10.1109/bigdata.2018.8622525

Cited by 271 publications

(189 citation statements)

References 17 publications

Supporting

Mentioning

189

Contrasting

Order By: Relevance

“…Generating data with unseen class, domain combi-nations The second family generates data samples associated with unseen class, domain combinations, such as ELEGANT [35], DNA-GAN [34], Multi-Level Variational Autoencoder (ML-VAE) [5], CausalGAN [14], Res-GAN [27], SaGAN [40], among others. FML methods Fairness GAN [26] and FairGAN [36] also fall into this category. These methods generate synthetic data, then ordinary models can be trained on both real and the generated data.…”

Section: Related Workmentioning

confidence: 99%

Additive Adversarial Learning for Unbiased Authentication

Liang

Cao

Zhang

et al. 2019

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

View full text Add to dashboard Cite

Authentication is a task aiming to confirm the truth between data instances and personal identities. Typical authentication applications include face recognition, person re-identification, authentication based on mobile devices and so on. The recently-emerging data-driven authentication process may encounter undesired biases, i.e., the models are often trained in one domain (e.g., for people wearing spring outfits) while required to apply in other domains (e.g., they change the clothes to summer outfits). To address this issue, we propose a novel two-stage method that disentangles the class/identity from domaindifferences, and we consider multiple types of domaindifference. In the first stage, we learn disentangled representations by a one-versus-rest disentangle learning (OVRDL) mechanism. In the second stage, we improve the disentanglement by an additive adversarial learning (AAL) mechanism. Moreover, we discuss the necessity to avoid a learning dilemma due to disentangling causally related types of domain-difference. Comprehensive evaluation results demonstrate the effectiveness and superiority of the proposed method.

show abstract

Section: Related Workmentioning

confidence: 99%

Additive Adversarial Learning for Unbiased Authentication

Liang

Cao

Zhang

et al. 2019

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

View full text Add to dashboard Cite

show abstract

“…Disparate treatment, a direct form of discrimination, results from a deliberated use of the sensitive attribute and can be avoided by removing it from the data prior to training the model [12]. Even when trained without the sensitive attribute, the predictions may still be discriminatory, leading to an unfair treatment of protected groups [12], [13]. This red-lining effect is due to the presence of features highly associated with the sensitive attribute [12], [13] and is linked to disparate impact.…”

Section: B Fairness Conceptsmentioning

confidence: 99%

“…Even when trained without the sensitive attribute, the predictions may still be discriminatory, leading to an unfair treatment of protected groups [12], [13]. This red-lining effect is due to the presence of features highly associated with the sensitive attribute [12], [13] and is linked to disparate impact. This indirect form of discrimination is not illegal in itself, as long as objective and reasonable justifications for it can be given [14], [15].…”

Section: B Fairness Conceptsmentioning

confidence: 99%

“…Statistical parity difference, or the Calders-Verwer score (CVS), considers the difference of the rate of favourable predictions between protected and unprotected groups [12]:…”

Section: B Fairness Conceptsmentioning

confidence: 99%

“…The approaches that have been proposed to enhance the fairness of ML models or mitigate the bias in their predictions can be grouped into three categories: pre-process, in-process and post-process. Pre-process approaches modify the training data to make it free of discrimination; in-process approaches change the models by adding constraints and regularisation terms to the objective functions; and post-process approaches directly change the predictions made by the models [12].…”

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

The Impact of Data Preparation on the Fairness of Software Systems

Valentim

Lourenço

Antunes

2019

2019 IEEE 30th International Symposium on Software Reliability Engineering (ISSRE)

View full text Add to dashboard Cite

Machine learning models are widely adopted in scenarios that directly affect people. The development of software systems based on these models raises societal and legal concerns, as their decisions may lead to the unfair treatment of individuals based on attributes like race or gender. Data preparation is key in any machine learning pipeline, but its effect on fairness is yet to be studied in detail. In this paper, we evaluate how the fairness and effectiveness of the learned models are affected by the removal of the sensitive attribute, the encoding of the categorical attributes, and instance selection methods (including cross-validators and random undersampling). We used the Adult Income and the German Credit Data datasets, which are widely studied and known to have fairness concerns. We applied each data preparation technique individually to analyse the difference in predictive performance and fairness, using statistical parity difference, disparate impact, and the normalised prejudice index. The results show that fairness is affected by transformations made to the training data, particularly in imbalanced datasets. Removing the sensitive attribute is insufficient to eliminate all the unfairness in the predictions, as expected, but it is key to achieve fairer models. Additionally, the standard random undersampling with respect to the true labels is sometimes more prejudicial than performing no random undersampling.

show abstract

Bias in data‐driven artificial intelligence systems—An introductory survey

Ntoutsi

Fafalios

Gadiraju

et al. 2020

WIREs Data Min & Knowl

474

259

View full text Add to dashboard Cite

Artificial Intelligence (AI)-based systems are widely employed nowadays to make decisions that have far-reaching impact on individuals and society. Their decisions might affect everyone, everywhere, and anytime, entailing concerns about potential human rights issues. Therefore, it is necessary to move beyond traditional AI algorithms optimized for predictive performance and embed ethical and legal principles in their design, training, and deployment to ensure social good while still benefiting from the huge potential of the AI technology. The goal of this survey is to provide a broad multidisciplinary overview of the area of bias in AI systems, focusing on technical challenges and solutions as well as to suggest new research directions towards approaches well-grounded in a legal frame. In this survey, we focus on data-driven AI, as a large part of AI is powered nowadays by (big) data and powerful machine learning algorithms. If otherwise not specified, we use the general term bias to describe problems related to the gathering or processing of data that might result in prejudiced decisions on the bases of demographic features such as race, sex, and so forth.This article is categorized under:

show abstract

FairGAN: Fairness-aware Generative Adversarial Networks

Cited by 271 publications

References 17 publications

Additive Adversarial Learning for Unbiased Authentication

Additive Adversarial Learning for Unbiased Authentication

The Impact of Data Preparation on the Fairness of Software Systems

Bias in data‐driven artificial intelligence systems—An introductory survey

Contact Info

Product

Resources

About