pith. sign in

arxiv: 2605.17118 · v1 · pith:ZXYLCUWMnew · submitted 2026-05-16 · 💻 cs.LG · stat.CO· stat.ML

Differentiable Optimization Layers for Guaranteed Fairness in Deep Learning

David Troxell , Noah Roemer , Guido Mont\'ufar This is my paper

Pith reviewed 2026-05-20 15:02 UTC · model grok-4.3

classification 💻 cs.LG stat.COstat.ML
keywords fairness layerdifferentiable optimizationneural networksoutput parityprimal-dual algorithmdeep learning fairnessstreaming predictions
0
0 comments X

The pith

A differentiable fairness layer appended to neural network outputs guarantees chosen notions of output parity.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proposes adding a fairness layer to the output of a neural network. This layer is a differentiable optimization module that enforces fairness constraints, such as output parity across groups, directly on the predictions. By integrating it into the model, fairness is satisfied by construction during both training and inference. An online primal-dual algorithm supports this in streaming settings with small batch sizes, providing provable guarantees. Sympathetic readers would value this because it embeds fairness into the learning process itself rather than relying on post-hoc adjustments.

Core claim

The authors introduce a fairness layer as a differentiable optimization layer appended to a model's output layer that guarantees a chosen notion of output parity is satisfied when integrated into a neural network. They also present an online primal-dual inference algorithm that provides provable aggregate fairness guarantees for streaming predictions with arbitrarily small batch sizes.

What carries the argument

The fairness layer, which formulates fairness as a convex optimization problem solved differentiably to project outputs onto the fair set.

If this is right

  • Neural networks equipped with the fairness layer produce predictions that exactly satisfy the selected fairness constraint.
  • The online primal-dual algorithm ensures aggregate fairness holds even when processing data in very small batches or streams.
  • Theoretical results characterize the differentiability and stability of the layer during backpropagation.
  • Empirical tests show the approach maintains model performance while achieving the fairness guarantees.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • This method might generalize to enforcing other types of constraints, such as monotonicity or robustness, in deep learning models.
  • Adopting the fairness layer could simplify compliance with fairness regulations by making guarantees part of the model architecture.
  • Future work could explore scaling the optimization layer to very large models or different fairness metrics.
  • Integration with existing fairness toolkits might allow hybrid approaches combining this with auditing methods.

Load-bearing premise

The fairness layer must remain differentiable and stable during backpropagation and model training for the guarantees to hold throughout the process.

What would settle it

Observing that predictions from a trained model with the fairness layer violate the output parity constraint on a test set would indicate the guarantee does not hold.

Figures

Figures reproduced from arXiv: 2605.17118 by David Troxell, Guido Mont\'ufar, Noah Roemer.

Figure 1
Figure 1. Figure 1: (a) ROC curves on the test set, averaged over 25 experiment repetitions with different, random train and test splits (with shaded regions representing ±1 standard deviation); (b) Precision-Recall curves on the test set, averaged over 25 experiment repeats (with shaded regions representing ±1 standard deviation); (c) Distribution of AUC and average precision on test set for all 25 experiment repetitions. me… view at source ↗
Figure 2
Figure 2. Figure 2: Test loss percent differences relative to F-Layer across 32 datasets. Each boxplot shows {(L (j) i − Fi)/Fi × 100} 32 i=1 where Fi is F-Layer loss and L (j) i is baseline j loss. Penalty model excludes cases where fairness constraints were violated [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Test loss rankings across all 32 dataset scenarios. to define the fairness layer. Here z = fθ(X(b) ) ∈ R nb . In other words, the constraint set ensures predictions stay within some bounds and Mean Conditional Parity (2) is sat￾isfied within a small tolerance ϵ = 0.05. See Appendix E for the corresponding Penalty and Strict Penalty formulations. The F-Layer, Projection, Penalty, and Strict Penalty models h… view at source ↗
Figure 4
Figure 4. Figure 4: Test loss rankings across all 32 dataset scenarios and batch size combinations. Black cells indicate constraint violations. yield modest accuracy gains over hard projection, consistent with softer enforcement allowing individual batches more expressivity. When stratified sampling is used, as in this ablation study, both regimes achieve aggregate fairness by Lemma 3.1. In streaming settings where stratified… view at source ↗
read the original abstract

Differentiable optimization layers are traditionally integrated in predict-then-optimize frameworks where a neural model estimates parameters that subsequently serve as fixed inputs to downstream decision-making optimization problems. In this work, we introduce the concept of a "fairness layer": a differentiable optimization layer appended to a model's output layer that guarantees a chosen notion of output parity is satisfied when integrated into a neural network. Additionally, we introduce an online primal-dual inference algorithm that provides provable aggregate fairness guarantees for streaming predictions with arbitrarily small batch sizes, where traditional per-batch constraints become overly restrictive. Numerical experiments demonstrate the effectiveness of the fairness layer and associated algorithm, and theoretical analysis characterizes the layer's differentiability and stability properties during model training and backpropagation. Our code for these experiments is publicly available on GitHub (https://github.com/dtroxell19/FairDL-ICML-2026.git) and our public Python package documentation can be found online: https://dtroxell19.github.io/fairness_training/.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces the concept of a fairness layer: a differentiable optimization layer appended to a neural network's output layer to guarantee a chosen notion of output parity. It also presents an online primal-dual inference algorithm that provides provable aggregate fairness guarantees for streaming predictions with arbitrarily small batch sizes. Theoretical analysis characterizes the layer's differentiability and stability during training and backpropagation, and numerical experiments demonstrate effectiveness. Public code and documentation are provided.

Significance. If the central claims hold, the work offers a practical way to enforce fairness constraints end-to-end within deep learning pipelines via differentiable optimization layers, addressing limitations of post-hoc or per-batch fairness methods. The streaming algorithm with small-batch guarantees is a notable contribution for real-world deployment. Public release of code strengthens reproducibility.

major comments (2)
  1. [§3.2] §3.2, the implicit-function-theorem argument for differentiability of the fairness layer: the required constraint qualification (e.g., LICQ or MFCQ) is not verified for the specific parity constraints used in the experiments; without it the solution map may fail to be differentiable at points encountered during training.
  2. [Theorem 4.1] Theorem 4.1 on aggregate fairness: the bound on cumulative fairness violation is stated to hold for arbitrarily small batch sizes, yet the proof sketch relies on a step-size schedule whose dependence on batch size is not made explicit, leaving open whether the guarantee remains non-vacuous for batch size 1.
minor comments (2)
  1. [Table 1] Table 1: the column headers for the fairness metrics are not aligned with the definitions given in §2.3; adding an explicit cross-reference would improve readability.
  2. [Figure 3] Figure 3 caption: the phrase 'fairness layer' is used without specifying which variant (hard vs. soft constraint) is plotted; this is a minor clarity issue.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading of the manuscript and the constructive comments. We appreciate the recommendation for minor revision and address each major comment below, making revisions to strengthen the theoretical presentation.

read point-by-point responses

  1. Referee: [§3.2] §3.2, the implicit-function-theorem argument for differentiability of the fairness layer: the required constraint qualification (e.g., LICQ or MFCQ) is not verified for the specific parity constraints used in the experiments; without it the solution map may fail to be differentiable at points encountered during training.

    Authors: We agree that an explicit verification of the constraint qualification strengthens the application of the implicit function theorem. The fairness constraints in our experiments are linear equality constraints (output parity across groups), whose gradients are linearly independent for any non-empty groups; thus LICQ holds at all feasible points. We will add a short paragraph in §3.2 stating this verification and noting that the solution map remains differentiable throughout training under the problem assumptions. The revision will be incorporated in the next version. revision: yes

  2. Referee: [Theorem 4.1] Theorem 4.1 on aggregate fairness: the bound on cumulative fairness violation is stated to hold for arbitrarily small batch sizes, yet the proof sketch relies on a step-size schedule whose dependence on batch size is not made explicit, leaving open whether the guarantee remains non-vacuous for batch size 1.

    Authors: We thank the referee for this observation. The step-size schedule used in the proof of Theorem 4.1 is independent of batch size B (specifically of the form 1/sqrt(t)), and the resulting O(sqrt(T)) bound on cumulative violation remains non-vacuous for B=1 because the constants do not diverge as B approaches 1. We will revise the proof sketch to make the lack of dependence on B explicit and add a remark confirming the guarantee for batch size 1. This clarification will be included in the revised manuscript. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected in derivation chain

full rationale

The paper introduces a new fairness layer construction as a differentiable optimization layer appended to a neural network's output to enforce a chosen output parity notion by design, along with an online primal-dual algorithm for aggregate fairness in streaming settings. Theoretical claims characterize differentiability, stability, and provable guarantees without reducing any central result to a self-citation chain, a fitted parameter renamed as a prediction, or a self-definitional loop where the output parity is presupposed in the layer's definition. The derivation remains self-contained as a novel integration of optimization layers with fairness constraints, supported by numerical experiments and public code rather than circular reductions to prior fitted quantities or author-specific uniqueness theorems.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available, so no specific free parameters, axioms, or invented entities can be extracted or audited from the full technical content.

pith-pipeline@v0.9.0 · 5708 in / 1087 out tokens · 70659 ms · 2026-05-20T15:02:07.773304+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

132 extracted references · 132 canonical work pages · 7 internal anchors

  1. [1]

    Health informatics: A computational perspective in healthcare , pages=

    Medical image generation using generative adversarial networks: A review , author=. Health informatics: A computational perspective in healthcare , pages=. 2021 , publisher=

    work page 2021

  2. [2]

    2018 IEEE 15th international symposium on biomedical imaging (ISBI 2018) , pages=

    GAN-based synthetic brain MR image generation , author=. 2018 IEEE 15th international symposium on biomedical imaging (ISBI 2018) , pages=. 2018 , organization=

    work page 2018

  3. [3]

    European Conference on Computer Vision , pages=

    Towards Reliable Advertising Image Generation Using Human Feedback , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    work page 2024

  4. [4]

    Companion Proceedings of the 26th International Conference on Intelligent User Interfaces , pages=

    Akin: Generating ui wireframes from ui design patterns using deep learning , author=. Companion Proceedings of the 26th International Conference on Intelligent User Interfaces , pages=

    work page

  5. [5]

    2021 IEEE/ACM 43rd International Conference on Software Engineering (ICSE) , pages=

    Guigan: Learning to generate gui designs using generative adversarial networks , author=. 2021 IEEE/ACM 43rd International Conference on Software Engineering (ICSE) , pages=. 2021 , organization=

    work page 2021

  6. [6]

    Electronic Commerce Research , volume=

    Ad creative generation using reinforced generative adversarial network , author=. Electronic Commerce Research , volume=. 2024 , publisher=

    work page 2024

  7. [7]

    arXiv preprint arXiv:2307.15326 , year=

    Staging e-commerce products for online advertising using retrieval assisted image generation , author=. arXiv preprint arXiv:2307.15326 , year=

    work page arXiv

  8. [8]

    Scientific Reports , volume=

    Denoising diffusion probabilistic models for 3D medical image generation , author=. Scientific Reports , volume=. 2023 , publisher=

    work page 2023

  9. [9]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Dall-eval: Probing the reasoning skills and social biases of text-to-image generation models , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    work page

  10. [10]

    Proceedings of the 32nd ACM International Conference on Multimedia , pages=

    New job, new gender? Measuring the social bias in image generation models , author=. Proceedings of the 32nd ACM International Conference on Multimedia , pages=

    work page

  11. [11]

    arXiv preprint arXiv:2308.00755 , year=

    The bias amplification paradox in text-to-image generation , author=. arXiv preprint arXiv:2308.00755 , year=

    work page arXiv

  12. [12]

    Proceedings of the 31st ACM International Conference on Multimedia , pages=

    Text-to-image diffusion models can be easily backdoored through multimodal data poisoning , author=. Proceedings of the 31st ACM International Conference on Multimedia , pages=

    work page

  13. [13]

    Generative Poisoning Attack Method Against Neural Networks

    Generative poisoning attack method against neural networks , author=. arXiv preprint arXiv:1703.01340 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  14. [14]

    Proceedings of the 2024 on ACM SIGSAC Conference on Computer and Communications Security , pages=

    Safegen: Mitigating sexually explicit content generation in text-to-image models , author=. Proceedings of the 2024 on ACM SIGSAC Conference on Computer and Communications Security , pages=

    work page 2024

  15. [15]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    High-resolution image synthesis with latent diffusion models , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    work page

  16. [16]

    Representation Learning with Contrastive Predictive Coding

    Representation learning with contrastive predictive coding , author=. arXiv preprint arXiv:1807.03748 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  17. [17]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Image generation from scene graphs , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    work page

  18. [18]

    Using Scene Graph Context to Improve Image Generation

    Using scene graph context to improve image generation , author=. arXiv preprint arXiv:1901.03762 , year=

    work page internal anchor Pith review Pith/arXiv arXiv 1901

  19. [19]

    Advances in Neural Information Processing Systems , volume=

    Pastegan: A semi-parametric method to generate image from scene graph , author=. Advances in Neural Information Processing Systems , volume=

    work page

  20. [20]

    Computer Vision--ECCV 2020: 16th European Conference, Glasgow, UK, August 23--28, 2020, Proceedings, Part XXVI 16 , pages=

    Learning canonical representations for scene graph to image generation , author=. Computer Vision--ECCV 2020: 16th European Conference, Glasgow, UK, August 23--28, 2020, Proceedings, Part XXVI 16 , pages=. 2020 , organization=

    work page 2020

  21. [21]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Scenegenie: Scene graph guided diffusion models for image synthesis , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    work page

  22. [22]

    Proceedings of the IEEE international conference on computer vision , pages=

    Photographic image synthesis with cascaded refinement networks , author=. Proceedings of the IEEE international conference on computer vision , pages=

    work page

  23. [23]

    , author=

    Image Synthesis with Graph Conditioning: CLIP-Guided Diffusion Models for Scene Graphs. , author=. CoRR , year=

    work page

  24. [24]

    International conference on machine learning , pages=

    Learning transferable visual models from natural language supervision , author=. International conference on machine learning , pages=. 2021 , organization=

    work page 2021

  25. [25]

    Advances in neural information processing systems , volume=

    Generative adversarial nets , author=. Advances in neural information processing systems , volume=

    work page

  26. [26]

    Advances in Neural Information Processing Systems , volume=

    Retrieval-augmented generation for knowledge-intensive nlp tasks , author=. Advances in Neural Information Processing Systems , volume=

    work page

  27. [27]

    A Deep Local and Global Scene-Graph Matching for Image-Text Retrieval , journal =

    Manh. A Deep Local and Global Scene-Graph Matching for Image-Text Retrieval , journal =. 2021 , url =. 2106.02400 , timestamp =

    work page arXiv 2021

  28. [28]

    arXiv preprint arXiv:2211.12561 , year=

    Retrieval-augmented multimodal language modeling , author=. arXiv preprint arXiv:2211.12561 , year=

    work page arXiv

  29. [29]

    CoRR , volume =

    Sijin Wang and Ruiping Wang and Ziwei Yao and Shiguang Shan and Xilin Chen , title =. CoRR , volume =. 2019 , url =. 1910.05134 , timestamp =

    work page arXiv 2019

  30. [30]

    Proceedings of the fourth workshop on vision and language , pages=

    Generating semantically precise scene graphs from textual descriptions for improved image retrieval , author=. Proceedings of the fourth workshop on vision and language , pages=

    work page

  31. [31]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Image retrieval using scene graphs , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    work page

  32. [32]

    2022 , eprint=

    Paint by Example: Exemplar-based Image Editing with Diffusion Models , author=. 2022 , eprint=

    work page 2022

  33. [33]

    arXiv preprint arXiv:2201.07520 , year=

    Cm3: A causal masked multimodal model of the internet , author=. arXiv preprint arXiv:2201.07520 , year=

    work page arXiv

  34. [34]

    Diffusion-Guided Counterfactual Generation for Model Explainability , author=

    work page

  35. [35]

    Proceedings of the Asian Conference on Computer Vision (ACCV) , month =

    Jeanneret, Guillaume and Simon, Loic and Jurie, Frederic , title =. Proceedings of the Asian Conference on Computer Vision (ACCV) , month =. 2022 , pages =

    work page 2022

  36. [36]

    arXiv preprint arXiv:2210.04885 , year=

    What the daam: Interpreting stable diffusion using cross attention , author=. arXiv preprint arXiv:2210.04885 , year=

    work page arXiv

  37. [37]

    2024 IEEE 18th International Conference on Automatic Face and Gesture Recognition (FG) , pages=

    Discovering interpretable directions in the semantic latent space of diffusion models , author=. 2024 IEEE 18th International Conference on Automatic Face and Gesture Recognition (FG) , pages=. 2024 , organization=

    work page 2024

  38. [38]

    Scene Graph Parsing as Dependency Parsing

    Scene graph parsing as dependency parsing , author=. arXiv preprint arXiv:1803.09189 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  39. [39]

    GPT-4 Technical Report

    Gpt-4 technical report , author=. arXiv preprint arXiv:2303.08774 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  40. [40]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Compositional chain-of-thought prompting for large multimodal models , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  41. [41]

    arXiv preprint arXiv:2405.15321 , year=

    SG-Adapter: Enhancing Text-to-Image Generation with Scene Graph Guidance , author=. arXiv preprint arXiv:2405.15321 , year=

    work page arXiv

  42. [42]

    arXiv preprint arXiv:2312.04314 , year=

    Gpt4sgg: Synthesizing scene graphs from holistic and region-specific narratives , author=. arXiv preprint arXiv:2312.04314 , year=

    work page arXiv

  43. [43]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    LLM4SGG: Large Language Models for Weakly Supervised Scene Graph Generation , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  44. [44]

    Scene-llm: Extending language model for 3d visual understanding and reasoning,

    Scene-llm: Extending language model for 3d visual understanding and reasoning , author=. arXiv preprint arXiv:2403.11401 , year=

    work page arXiv

  45. [45]

    arXiv preprint arXiv:2305.17497 , year=

    Factual: A benchmark for faithful and consistent textual scene graph parsing , author=. arXiv preprint arXiv:2305.17497 , year=

    work page arXiv

  46. [46]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Unified visual-semantic embeddings: Bridging vision and language with structured meaning representations , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    work page

  47. [47]

    Medical image computing and computer-assisted intervention--MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18 , pages=

    U-net: Convolutional networks for biomedical image segmentation , author=. Medical image computing and computer-assisted intervention--MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18 , pages=. 2015 , organization=

    work page 2015

  48. [48]

    Computer Vision--ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13 , pages=

    Microsoft coco: Common objects in context , author=. Computer Vision--ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13 , pages=. 2014 , organization=

    work page 2014

  49. [49]

    International journal of computer vision , volume=

    Visual genome: Connecting language and vision using crowdsourced dense image annotations , author=. International journal of computer vision , volume=. 2017 , publisher=

    work page 2017

  50. [50]

    Computational economics , volume=

    Credit risk assessment using statistical and machine learning: basic methodology and risk modeling applications , author=. Computational economics , volume=. 2000 , publisher=

    work page 2000

  51. [51]

    , author=

    When the Machine Meets the Expert: An Ethnography of Developing AI for Hiring. , author=. MIS quarterly , volume=

    work page

  52. [52]

    Journal of Experimental Criminology , volume=

    An impact assessment of machine learning risk forecasts on parole board decisions and recidivism , author=. Journal of Experimental Criminology , volume=. 2017 , publisher=

    work page 2017

  53. [53]

    Array , volume=

    Deep learning for object detection and scene perception in self-driving cars: Survey, challenges, and open issues , author=. Array , volume=. 2021 , publisher=

    work page 2021

  54. [54]

    Multimodal Technologies and Interaction , volume=

    Deep learning and medical diagnosis: A review of literature , author=. Multimodal Technologies and Interaction , volume=. 2018 , publisher=

    work page 2018

  55. [55]

    2018 , isbn =

    Verma, Sahil and Rubin, Julia , title =. 2018 , isbn =. doi:10.1145/3194770.3194776 , booktitle =

    work page doi:10.1145/3194770.3194776 2018

  56. [56]

    Proceedings of the 30th International Conference on Neural Information Processing Systems , pages =

    Hardt, Moritz and Price, Eric and Srebro, Nathan , title =. Proceedings of the 30th International Conference on Neural Information Processing Systems , pages =. 2016 , publisher =

    work page 2016

  57. [57]

    Proceedings of the 3rd Innovations in Theoretical Computer Science Conference , pages =

    Dwork, Cynthia and Hardt, Moritz and Pitassi, Toniann and Reingold, Omer and Zemel, Richard , title =. Proceedings of the 3rd Innovations in Theoretical Computer Science Conference , pages =. 2012 , noisbn =

    work page 2012

  58. [58]

    Proceedings of the 37th International Conference on Machine Learning , articleno =

    Mukherjee, Debarghya and Yurochkin, Mikhail and Banerjee, Moulinath and Sun, Yuekai , title =. Proceedings of the 37th International Conference on Machine Learning , articleno =. 2020 , publisher =

    work page 2020

  59. [59]

    2019 , publisher =

    Adel, Tameem and Valera, Isabel and Ghahramani, Zoubin and Weller, Adrian , title =. 2019 , publisher =. doi:10.1609/aaai.v33i01.33012412 , booktitle =

    work page doi:10.1609/aaai.v33i01.33012412 2019

  60. [60]

    Ethical Adversaries: Towards Mitigating Unfairness with Adversarial Machine Learning , year =

    Delobelle, Pieter and Temple, Paul and Perrouin, Gilles and Fr\'. Ethical Adversaries: Towards Mitigating Unfairness with Adversarial Machine Learning , year =. doi:10.1145/3468507.3468513 , journal =

    work page doi:10.1145/3468507.3468513

  61. [61]

    International Conference on Learning Representations , year=

    Mode Regularized Generative Adversarial Networks , author=. International Conference on Learning Representations , year=

    work page

  62. [62]

    2021 , booktitle =

    Padala, Manisha and Gujar, Sujit , title =. 2021 , booktitle =

    work page 2021

  63. [63]

    Journal of Machine Learning Research , year =

    Andrew Cotter and Heinrich Jiang and Maya Gupta and Serena Wang and Taman Narayan and Seungil You and Karthik Sridharan , title =. Journal of Machine Learning Research , year =

    work page

  64. [64]

    , booktitle =

    Zafar, Muhammad Bilal and Valera, Isabel and Rogriguez, Manuel Gomez and Gummadi, Krishna P. , booktitle =. 2017 , noeditor =

    work page 2017

  65. [65]

    Elisa and Huang, Lingxiao and Keswani, Vijay and Vishnoi, Nisheeth K

    Celis, L. Elisa and Huang, Lingxiao and Keswani, Vijay and Vishnoi, Nisheeth K. , title =. Proceedings of the Conference on Fairness, Accountability, and Transparency , pages =. 2019 , noisbn =

    work page 2019

  66. [66]

    2004 , publisher=

    Convex optimization , author=. 2004 , publisher=

    work page 2004

  67. [67]

    Gummadi , title =

    Muhammad Bilal Zafar and Isabel Valera and Manuel Gomez-Rodriguez and Krishna P. Gummadi , title =. Journal of Machine Learning Research , year =

    work page

  68. [68]

    The World Wide Web Conference , pages =

    Wu, Yongkai and Zhang, Lu and Wu, Xintao , title =. The World Wide Web Conference , pages =. 2019 , noisbn =

    work page 2019

  69. [69]

    Zafar, Muhammad Bilal and Valera, Isabel and Gomez Rodriguez, Manuel and Gummadi, Krishna P. , year=. Fairness Beyond Disparate Treatment and Disparate Impact: Learning Classification without Disparate Mistreatment , DOI=. Proceedings of the 26th International Conference on World Wide Web , publisher=

    work page

  70. [70]

    and Weller, Adrian , title =

    Zafar, Muhammad Bilal and Valera, Isabel and Rodriguez, Manuel Gomez and Gummadi, Krishna P. and Weller, Adrian , title =. Advances in Neural Information Processing Systems 30 (NeurIPS 2017) , year =

    work page 2017

  71. [71]

    2018 , isbn =

    Goel, Naman and Yaghini, Mohammad and Faltings, Boi , title =. 2018 , isbn =. doi:10.1145/3278721.3278722 , booktitle =

    work page doi:10.1145/3278721.3278722 2018

  72. [72]

    Proceedings of the 36th International Conference on Machine Learning , pages =

    Fairness risk measures , author =. Proceedings of the 36th International Conference on Machine Learning , pages =. 2019 , editor =

    work page 2019

  73. [73]

    Deep Declarative Networks , DOI=

    Gould, Stephen and Hartley, Richard and Campbell, Dylan John , year=. Deep Declarative Networks , DOI=. IEEE Transactions on Pattern Analysis and Machine Intelligence , publisher=

    work page

  74. [74]

    Karush, William , biburl =

    work page

  75. [75]

    Kuhn, H. W. and Tucker, A. W. , biburl =. Proceedings of the

    work page

  76. [76]

    Global optimization: From theory to implementation , pages=

    Disciplined convex programming , author=. Global optimization: From theory to implementation , pages=. 2006 , publisher=

    work page 2006

  77. [77]

    Zico , title =

    Agrawal, Akshay and Amos, Brandon and Barratt, Shane and Boyd, Stephen and Diamond, Steven and Kolter, J. Zico , title =. Advances in Neural Information Processing Systems 32 (NeurIPS 2019) , year =

    work page 2019

  78. [78]

    and Boyd, Stephen P

    Grant, Michael C. and Boyd, Stephen P. Graph Implementations for Nonsmooth Convex Programs. Recent Advances in Learning and Control. 2008

    work page 2008

  79. [79]

    Grant, Michael and Boyd, Stephen , year=

    work page

  80. [80]

    Computational Optimization and Applications , year=

    Solution refinement at regular points of conic problems , author=. Computational Optimization and Applications , year=

    work page

Showing first 80 references.