FairGAN: Fairness-aware Generative Adversarial Networks

By Depeng Xu, Shuhan Yuan, Lu Zhang, Xintao Wu. 2018

379 citations recorded by the source corpus.

Fairness-aware learning is increasingly important in data mining. Discrimination prevention aims to prevent discrimination in the training data before it is used to conduct predictive analysis. In this paper, we focus on fair data generation that ensures the generated data is discrimination free. Inspired by generative adversarial networks (GAN), we present fairness-aware generative adversarial networks, called FairGAN, which are able to learn a generator producing fair data and also preserving good data utility. Compared with the naive fair data generation models, FairGAN further ensures the classifiers which are trained on generated data can achieve fair classification on real data. Experiments on a real dataset show the effectiveness of FairGAN.

Subfields: Artificial Intelligence, Safety Research, Generative Models, AI Ethics, Fairness & Society, Philosophy

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Safety Research

In Generative Models

In AI Ethics, Fairness & Society

In Philosophy

Citations in the corpus

A Survey on Bias and Fairness in Machine Learning (2019, methodology): “Structure of FairGAN as proposed in [129].”
Algorithmic Fairness (2020, methodology): “[139], for example, use a GAN framework with one generator and two discriminators.”
Learning Generative Adversarial RePresentations (GAP) under Fairness and Censoring Constraints (2019, background): “Using GANs to generate synthetic non-sensitive attributes and labels which ensure fairness while preserving the utility of the data (predicting the label) has been studied in (Xu et al., 2018; Sattigeri et al., 2018).”
Pros and Cons of GAN Evaluation Measures: New Developments (2021, background): “Some recent works ( e.g. Xu et al. (2018); Yu et al. (2020) have tried to mitigate the bias in GANs or use GANs to reduce bias in other ML algorithms ( e.g. Sattigeri et al. (2019); McDuﬀ et al. (2019)).”
Avoiding Resentment Via Monotonic Fairness (2019, methodology): “Neural networks have been used to learn fair classifiers in a number of contexts [33, 6, 35, 42].”
The ethics of algorithms: key problems and solutions (2020, neutral): “It is also worth noting that so-called ‘synthetic data’, or algorithmically generated data, produced via reinforcement learning or generative adversarial networks (GANs) offer an opportunity to address certain issues of data bias (Floridi 2019a; Xu et al. 2018).”
Machine Learning for Synthetic Data Generation: a Review (2023, background): “Authors in [200] leverage GAN as the pre-processing for fair data generation that ensures the generated data is discrimination free while maintaining high data utility.”
Fairness in Machine Learning: A Survey (2020, neutral): “[277, 194, 97, 278, 3, 152]) with similar objectives to transformation as discussed in subsection 4.”
Rényi Fair Inference (2019, methodology): “…to fairness in machine learning in recent years; see Calmon et al. (2017); Feldman et al. (2015); Hardt et al. (2016); Zhang et al. (2018); Xu et al. (2018); Dwork et al. (2018); Fish et al. (2016); Woodworth et al. (2017); Zafar et al. (2017; 2015); P´erez-Suay et al. (2017); Bechavod &…”
Copyright in generative deep learning (2021, neutral): “…of machine learning researches related to fairness, see Chouldechova and Roth (2020)) because of biases and discrimination they may cause (as happened with AI Portrait Ars [O ’ Leary, 2019], leading to some very remarkable attempts to overcome them, as in Xu et al. (2018) or Yu et al. (2020)).”
Debiasing Pre-Trained Language Models via Efficient Fine-Tuning (2022, methodology): “To mitigate biased behaviors in machine learning models, researchers have proposed meth-ods targeting different tasks and domains, such as classification (Menon and Williamson, 2018; Roh et al., 2021), regression (Agarwal et al., 2019; Berk et al., 2017), and adversarial learning (Xu et al., 2018).”
Fair Generative Modeling via Weak Supervision (2019, background): “D.1 Downstream Classification Task We note that although it is difficult to directly compare our model to supervised baselines such as FairGAN [39] and FairnessGAN [40] due to the unsupervised nature of our work, we conduct further evaluations on a relevant downstream task classification task, adapt”
Federated Machine Learning: Survey, Multi-Level Classification, Desirable Criteria and Future Directions in Communication and Networking Systems (2021, background): “Although many approaches have been lately proposed to fight against bias in traditional machine learning systems [86], [87], this topic has”
Inclusive GAN: Improving Data and Minority Coverage in Generative Models (2020, background): “However, these advances have come at a cost: there could be potential biases in the learned model against underrepresented data subgroups [49,57,45,16,17].”
ContraGAN: Contrastive Learning for Conditional Image Generation (2020, background): “This can trigger sexual harassment [60], fake news [61], and even security issues of face recognition systems [62].”
Data collection and quality challenges in deep learning: a data-centric AI perspective (2021, methodology): “Model patching [65] utilizes GANs to augment the data of speciﬁc subgroups of a class so that the model accuracy is similar across subgroups.”
A Review on Generative Adversarial Networks: Algorithms, Theory, and Applications (2020, background): “[309], [310], [311], [312], [313], [314], [315], [316], to generate neural networks [317], to augment data [318], [319], to learn spatial representations [320], and in network embedding [321], heterogeneous information networks [322], and”