pith. sign in

arxiv: 2605.20975 · v2 · pith:T6M67VKJnew · submitted 2026-05-20 · 💻 cs.LG · cs.CR

Choose Wisely and Privately: Proactive Client Selection for Fair and Efficient Federated Learning

Adda Akram Bendoukha , Heber Hwang Arcolezi , Nesrine Kaaniche , Aymen Boudguiga This is my paper

Pith reviewed 2026-05-22 10:06 UTC · model grok-4.3

classification 💻 cs.LG cs.CR
keywords federated learningclient selectiondifferential privacyfairnessmutual informationsimulated annealingnon-IID datacontingency tables
0
0 comments X

The pith

Selecting clients proactively with mutual information from private tables yields faster, fairer, and more accurate federated models than uniform or adaptive sampling.

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

The paper proposes a proactive client selection framework that identifies an optimal fixed-size federation before training begins. It quantifies the relevance of cross-feature correlations in the union of client datasets using mutual information derived from differentially private contingency tables. A Potential Federation Loss balances collective data utility against fair cross-feature correlations, and this objective is optimized as a subset search problem via simulated annealing under differential privacy. Experiments across four benchmarks show that models trained on these selected federations converge faster, reach higher accuracy, and exhibit better group fairness than those produced by uniform sampling or existing adaptive aggregation strategies.

Core claim

An optimal federation found by minimizing Potential Federation Loss over mutual information from differentially private client statistics produces federated models that train faster, achieve higher final accuracy, and reduce group unfairness compared with uniform or state-of-the-art reactive client sampling.

What carries the argument

The Potential Federation Loss (PFL), a scalar objective that trades off aggregate data utility against cross-feature correlation fairness, optimized by simulated annealing on differentially private contingency tables.

If this is right

  • Fewer communication rounds are needed to reach target accuracy because low-utility clients are excluded upfront.
  • Total privacy exposure decreases since only the selected clients upload gradients during training.
  • Reactive mechanisms that discard or heavily down-weight client contributions become unnecessary.
  • Group-level unfairness arising from heterogeneous feature correlations is reduced by explicit fairness terms in the selection objective.

Where Pith is reading between the lines

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

  • The same pre-training selection logic could be applied periodically in long-running federated systems whose data distributions drift.
  • Energy and bandwidth savings at edge devices would follow from skipping gradient computations on clients outside the chosen federation.
  • The approach might generalize to other collaborative settings that face non-IID data and privacy constraints, such as distributed analytics or multi-party computation.

Load-bearing premise

Mutual information computed from differentially private contingency tables sufficiently captures both collective data utility and cross-feature fairness so that optimizing the Potential Federation Loss produces federations whose downstream models improve performance without excessive noise or search cost.

What would settle it

On the same four benchmarks, training models on the PFL-selected federations yields no statistically significant gains in convergence rounds, test accuracy, or fairness metrics relative to uniform sampling or current adaptive baselines.

Figures

Figures reproduced from arXiv: 2605.20975 by Adda Akram Bendoukha, Aymen Boudguiga, Heber Hwang Arcolezi, Nesrine Kaaniche.

Figure 1
Figure 1. Figure 1: Overview of our approach in three steps. First step involves computing cross-features contingency tables, and adding [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Correlations between the PFL and utility/fairness met￾rics using weights found through the aforementioned process. computable from local statistics. However, we note that these values act as domain-specific proxies rather than universal constants. Because MI inherently scales with feature en￾tropy, and the fairness-utility trade-off varies across different empirical data, applying our framework to highly d… view at source ↗
Figure 3
Figure 3. Figure 3: Distribution of MI estimation error under Gaussian [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Convergence and fairness of our optimal federations compared with random ones, with FedAvg, FedProx, and [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Convergence and fairness of our optimal federations compared with random ones, with sampling strategies: UCB-CS [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Privacy auditing using Log-likelihood Ratio Test (LRT) [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Convergence and fairness of our optimal federations [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Convergence and fairness of our optimal federations [PITH_FULL_IMAGE:figures/full_fig_p018_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Convergence and fairness of our optimal federations compared with random ones, with FedAvg, FedProx, and [PITH_FULL_IMAGE:figures/full_fig_p019_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Convergence and fairness of our optimal federations compared with random ones, with sampling strategies: UCB-CS [PITH_FULL_IMAGE:figures/full_fig_p019_10.png] view at source ↗
read the original abstract

Federated Learning enables collaborative model training across decentralized data sources without data transfer. Averaging-based FL is limited by the presence of non-IID data, which negatively impacts convergence speed and final model accuracy. Conventional alternatives suffer from significant inefficiency. Clients with noisy or highly heterogeneous data contribute expensive gradient computations that are either discarded or heavily down-weighted before aggregation. These reactive approaches waste computational resources, require more communication rounds and result in unnecessary privacy exposure. In this paper, we propose a proactive client selection framework that aims to find an optimal federation of clients whose combined data match utility and fairness requirements before training begins. Our method relies on mutual information computed from differentially private contingency tables to quantify the relevance of cross-feature correlations in the union dataset. We introduce a Potential Federation Loss (PFL) over the set of fixed-size federations, which balances two objectives. Maximizing collective data utility while ensuring fair cross-features correlations to prevent group unfairness. Client selection is expressed as an optimal subset search problem over the PFL objective, which we solve using simulated annealing under strong differential privacy guarantees for clients' local statistics. Experimental results on four benchmarks show faster, fairer, and more accurate models trained on optimally found federations, compared to uniform sampling, even when state-of-the-art adaptive aggregation or sampling strategies are employed.

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 proposes a proactive client selection framework for federated learning. It computes mutual information from differentially private contingency tables to quantify cross-feature correlations in the union of client data. A Potential Federation Loss (PFL) is introduced to balance collective data utility against fair cross-feature correlations. Client selection is formulated as a fixed-size subset search problem solved via simulated annealing under differential privacy. The central claim is that models trained on the resulting federations converge faster, achieve higher accuracy, and exhibit better fairness than those from uniform sampling or state-of-the-art adaptive aggregation/sampling strategies, as demonstrated on four benchmarks.

Significance. If the empirical results hold, the work would offer a meaningful advance by shifting from reactive client weighting or discarding to proactive, privacy-aware selection that reduces wasted computation and communication while addressing non-IID challenges and group fairness. The introduction of PFL as an explicit utility-fairness objective and the application of simulated annealing to the selection problem under DP guarantees are technically interesting contributions that could influence practical FL system design.

major comments (2)
  1. [Experimental results and method formulation] The central experimental claim (faster, fairer, more accurate models on four benchmarks) rests on the assumption that mutual information derived from DP contingency tables, when minimized via PFL and simulated annealing, selects federations whose union data genuinely improves downstream performance. However, no analysis is provided on how Laplace or Gaussian noise (controlled by epsilon) affects MI estimates in non-IID or small-client regimes typical of FL benchmarks; moderate privacy budgets can flatten correlations and produce noisy or unrepresentative selections whose reported gains may be illusory.
  2. [Abstract and experimental evaluation] The abstract states improvements even against SOTA adaptive aggregation or sampling strategies, yet provides no details on experimental setup, exact baselines, statistical significance testing, federation size k, privacy budget epsilon values, or how PFL is precisely formulated and optimized. This absence makes it impossible to verify whether the gains are robust or load-bearing for the proactive-selection thesis.
minor comments (2)
  1. [Method] Notation for the PFL objective and the contingency-table construction should be introduced with explicit equations early in the method section to improve readability.
  2. [Experimental results] The paper should clarify whether the four benchmarks are standard FL datasets (e.g., CIFAR-10, FEMNIST) and report client counts, data heterogeneity measures, and number of independent runs.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback and for recognizing the potential significance of our proactive client selection framework. We address each major comment below, clarifying aspects of our method and experiments while outlining targeted revisions to improve clarity, robustness, and reproducibility.

read point-by-point responses

  1. Referee: [Experimental results and method formulation] The central experimental claim (faster, fairer, more accurate models on four benchmarks) rests on the assumption that mutual information derived from DP contingency tables, when minimized via PFL and simulated annealing, selects federations whose union data genuinely improves downstream performance. However, no analysis is provided on how Laplace or Gaussian noise (controlled by epsilon) affects MI estimates in non-IID or small-client regimes typical of FL benchmarks; moderate privacy budgets can flatten correlations and produce noisy or unrepresentative selections whose reported gains may be illusory.

    Authors: We agree that a dedicated sensitivity analysis of the impact of differential privacy noise on mutual information estimates would strengthen the empirical claims, particularly for non-IID distributions and smaller client pools. Our current results already evaluate performance across a range of privacy budgets, but we did not include an explicit breakdown of how noise perturbs the contingency tables and resulting MI values. In the revised manuscript we will add new experiments that (i) vary epsilon over a fine-grained grid, (ii) quantify the deviation between noisy and non-private MI estimates, and (iii) report the stability of the selected client sets and downstream accuracy under these conditions. This will directly address whether moderate privacy budgets produce unrepresentative selections. revision: yes

  2. Referee: [Abstract and experimental evaluation] The abstract states improvements even against SOTA adaptive aggregation or sampling strategies, yet provides no details on experimental setup, exact baselines, statistical significance testing, federation size k, privacy budget epsilon values, or how PFL is precisely formulated and optimized. This absence makes it impossible to verify whether the gains are robust or load-bearing for the proactive-selection thesis.

    Authors: The full experimental protocol—including federation size k, concrete epsilon values, the exact mathematical formulation and simulated-annealing optimization of PFL, the complete list of baselines with citations, and the statistical testing procedure—is presented in Sections 4 (Experimental Setup) and 5 (Results). Nevertheless, we acknowledge that the abstract is too high-level for immediate verification. We will revise the abstract to mention key parameter ranges and will add a concise experimental-configuration table plus explicit reporting of statistical significance (paired t-tests or Wilcoxon signed-rank tests with p-values) to the results section. These changes will make the reproducibility information readily accessible without altering the manuscript’s core claims. revision: partial

Circularity Check

0 steps flagged

No significant circularity; PFL defined independently as new objective

full rationale

The paper introduces Potential Federation Loss (PFL) explicitly as a novel objective that balances collective data utility (via mutual information on DP contingency tables) against cross-feature fairness. Client selection is framed as a standard combinatorial optimization problem solved by simulated annealing under DP guarantees. Experimental claims rest on benchmark comparisons to external baselines (uniform sampling, adaptive aggregation), not on any reduction of outputs to fitted inputs or self-referential definitions by construction. No load-bearing derivation step collapses to tautology or prior self-citation chains; the method is self-contained against the stated benchmarks.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 1 invented entities

The approach rests on the assumption that private mutual information from contingency tables can proxy data utility and fairness, plus standard optimization assumptions for simulated annealing.

free parameters (2)
  • federation size k
    Search is performed over fixed-size subsets of clients.
  • privacy budget epsilon
    Controls noise added to contingency tables for differential privacy.
axioms (1)
  • domain assumption Mutual information from cross-feature correlations in the union dataset quantifies relevance for utility and fairness
    Invoked to define the PFL objective before training.
invented entities (1)
  • Potential Federation Loss (PFL) no independent evidence
    purpose: Objective function balancing collective data utility and fair cross-feature correlations
    New loss introduced to guide client selection; no independent evidence provided.

pith-pipeline@v0.9.0 · 5781 in / 1279 out tokens · 37280 ms · 2026-05-22T10:06:51.233405+00:00 · methodology

Review history (2 revisions) →

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

59 extracted references · 59 canonical work pages · 2 internal anchors

  1. [1]

    Communication-efficient learning of deep networks from decentralized data,

    B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas, “Communication-efficient learning of deep networks from decentralized data,” inArtificial intelligence and statistics. Pmlr, 2017, pp. 1273– 1282

    work page 2017

  2. [2]

    Advances and open problems in federated learning,

    P. Kairouz, H. B. McMahan, B. Avent, A. Bellet, M. Bennis, A. N. Bhagoji, K. Bonawitz, Z. Charles, G. Cormode, R. Cummings, R. G. L. D’Oliveira, H. Eichner, S. E. Rouayheb, D. Evans, J. Gardner, Z. Garrett, A. Gasc ´on, B. Ghazi, P. B. Gibbons, M. Gruteser, Z. Harchaoui, C. He, L. He, Z. Huo, B. Hutchinson, J. Hsu, M. Jaggi, T. Javidi, G. Joshi, M. Khodak...

    work page arXiv 2021

  3. [3]

    Federated optimization in heterogeneous networks,

    T. Li, A. K. Sahu, M. Zaheer, M. Sanjabi, A. Talwalkar, and V . Smith, “Federated optimization in heterogeneous networks,” 2020. [Online]. Available: https://arxiv.org/abs/1812.06127

    work page arXiv 2020

  4. [4]

    Tackling the objective inconsistency problem in heterogeneous federated optimization,

    J. Wang, Q. Liu, H. Liang, G. Joshi, and H. V . Poor, “Tackling the objective inconsistency problem in heterogeneous federated optimization,” 2020. [Online]. Available: https://arxiv.org/abs/2007. 07481

    work page 2020

  5. [6]

    Bias propagation in federated learning,

    H. Chang and R. Shokri, “Bias propagation in federated learning,”arXiv preprint arXiv:2309.02160, 2023

    work page arXiv 2023

  6. [7]

    Federated learning with partial model personalization,

    K. Pillutla, K. Malik, A.-R. Mohamed, M. Rabbat, M. Sanjabi, and L. Xiao, “Federated learning with partial model personalization,” in Proceedings of the 39th International Conference on Machine Learning, ser. Proceedings of Machine Learning Research, K. Chaudhuri, S. Jegelka, L. Song, C. Szepesvari, G. Niu, and S. Sabato, Eds., vol

    work page

  7. [8]

    17 716–17 758

    PMLR, 17–23 Jul 2022, pp. 17 716–17 758. [Online]. Available: https://proceedings.mlr.press/v162/pillutla22a.html

    work page 2022

  8. [9]

    Machine learning with adversaries: Byzantine tolerant gradient descent,

    P. Blanchard, E. M. El Mhamdi, R. Guerraoui, and J. Stainer, “Machine learning with adversaries: Byzantine tolerant gradient descent,” inAdvances in Neural Information Processing Systems, I. Guyon, U. V . Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, Eds., vol. 30. Curran Associates, Inc., 2017. [Online]. Available: https://p...

    work page 2017

  9. [10]

    Fairfed: Enabling group fairness in federated learning,

    Y . H. Ezzeldin, S. Yan, C. He, E. Ferrara, and A. S. Avestimehr, “Fairfed: Enabling group fairness in federated learning,” inProceedings of the AAAI conference on artificial intelligence, vol. 37, no. 6, 2023, pp. 7494– 7502

    work page 2023

  10. [11]

    Fade: Federated aggregation with discrim- ination elimination,

    A.-A. Bendoukha, H. H. Arcolezi, N. Kaaniche, A. Boudguiga, R. Sirdey, and P.-E. Clet, “Fade: Federated aggregation with discrim- ination elimination,” inProceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency, ser. FAccT ’25. New York, NY , USA: Association for Computing Machinery, 2025, p. 3182–3195

    work page 2025

  11. [12]

    Improving fairness via federated learning,

    Y . Zeng, H. Chen, and K. Lee, “Improving fairness via federated learning,”arXiv preprint arXiv:2110.15545, 2021

    work page arXiv 2021

  12. [13]

    Sok: On gradient leakage in federated learning,

    J. Du, J. Hu, Z. Wang, P. Sun, N. Z. Gong, K. Ren, and C. Chen, “Sok: On gradient leakage in federated learning,”arXiv preprint arXiv:2404.05403, 2024

    work page arXiv 2024

  13. [14]

    Exploiting unintended feature leakage in collaborative learning,

    L. Melis, C. Song, E. De Cristofaro, and V . Shmatikov, “Exploiting unintended feature leakage in collaborative learning,” in2019 IEEE symposium on security and privacy (SP). IEEE, 2019, pp. 691–706

    work page 2019

  14. [15]

    A survey of privacy attacks in machine learning,

    M. Rigaki and S. Garcia, “A survey of privacy attacks in machine learning,”ACM Computing Surveys, vol. 56, no. 4, p. 1–34, Nov. 2023. [Online]. Available: http://dx.doi.org/10.1145/3624010

    work page doi:10.1145/3624010 2023

  15. [16]

    Poseidon: Privacy-preserving federated neural network learning,

    S. Sav, A. Pyrgelis, J. R. Troncoso-Pastoriza, D. Froelicher, J.-P. Bossuat, J. S. Sousa, and J.-P. Hubaux, “Poseidon: Privacy-preserving federated neural network learning,”arXiv preprint arXiv:2009.00349, 2020

    work page arXiv 2009

  16. [17]

    Towards practical homomorphic aggregation in byzantine- resilient distributed learning,

    A. Choffrut, R. Guerraoui, R. Pinot, R. Sirdey, J. Stephan, and M. Zuber, “Towards practical homomorphic aggregation in byzantine- resilient distributed learning,” inProceedings of the 25th International Middleware Conference, ser. Middleware ’24. New York, NY , USA: Association for Computing Machinery, 2024, p. 431–444. [Online]. Available: https://doi.o...

    work page doi:10.1145/3652892.3700783 2024

  17. [18]

    Towards privacy-preserving and fairness-aware federated learning framework,

    A.-A. Bendoukha, D. Demirag, N. Kaaniche, A. Boudguiga, R. Sirdey, and S. Gambs, “Towards privacy-preserving and fairness-aware federated learning framework,” inPETs 2025 : Privacy Enhancing Technologies, vol. 2025, no. 1, Washinghton, DC, United States, Jul. 2025, pp. 845–865. [Online]. Available: https://hal.science/hal-04782394

    work page 2025

  18. [19]

    Un- veiling the (in)security of threshold fhe-based federated learning: The practical impact of recent cpad attacks,

    A.-A. Bendoukha, R. Sirdey, A. Boudguiga, and N. Kaaniche, “Un- veiling the (in)security of threshold fhe-based federated learning: The practical impact of recent cpad attacks,” in2025 IEEE 38th Computer Security Foundations Symposium (CSF), 2025, pp. 425–440

    work page 2025

  19. [20]

    Practical Secure Aggregation for Federated Learning on User-Held Data

    K. Bonawitz, V . Ivanov, B. Kreuter, A. Marcedone, H. B. McMa- han, S. Patel, D. Ramage, A. Segal, and K. Seth, “Practical secure aggregation for federated learning on user-held data,”arXiv preprint arXiv:1611.04482, 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  20. [21]

    An introduction to variable and feature selection,

    I. Guyon and A. Elisseeff, “An introduction to variable and feature selection,”J. Mach. Learn. Res., vol. 3, no. null, p. 1157–1182, Mar. 2003

    work page 2003

  21. [22]

    Data-centric artificial intelligence: A survey,

    D. Zha, Z. P. Bhat, K.-H. Lai, F. Yang, Z. Jiang, S. Zhong, and X. Hu, “Data-centric artificial intelligence: A survey,”ACM Computing Surveys, vol. 57, no. 5, pp. 1–42, 2025

    work page 2025

  22. [23]

    Unbiased look at dataset bias,

    A. Torralba and A. A. Efros, “Unbiased look at dataset bias,” inCVPR 2011, 2011, pp. 1521–1528

    work page 2011

  23. [24]

    Using mutual information for selecting features in supervised neural net learning,

    R. Battiti, “Using mutual information for selecting features in supervised neural net learning,”IEEE Transactions on Neural Networks, vol. 5, no. 4, pp. 537–550, 1994

    work page 1994

  24. [26]

    Fast binary feature selection with conditional mutual infor- mation,

    F. Fleuret, “Fast binary feature selection with conditional mutual infor- mation,”J. Mach. Learn. Res., vol. 5, p. 1531–1555, Dec. 2004

    work page 2004

  25. [27]

    Fairness through awareness,

    C. Dwork, M. Hardt, T. Pitassi, O. Reingold, and R. Zemel, “Fairness through awareness,” inProceedings of the 3rd innovations in theoretical computer science conference, 2012, pp. 214–226

    work page 2012

  26. [28]

    Equality of opportunity in supervised learning,

    M. Hardt, E. Price, and N. Srebro, “Equality of opportunity in supervised learning,”Advances in neural information processing systems, vol. 29, 2016

    work page 2016

  27. [29]

    Retiring adult: New datasets for fair machine learning,

    F. Ding, M. Hardt, J. Miller, and L. Schmidt, “Retiring adult: New datasets for fair machine learning,”Advances in neural information processing systems, vol. 34, pp. 6478–6490, 2021

    work page 2021

  28. [30]

    Communication-efficient learning of deep networks from decentralized data,

    B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas, “Communication-efficient learning of deep networks from decentralized data,” inAISTATS, 2017

    work page 2017

  29. [31]

    Calibrating noise to sensitivity in private data analysis,

    C. Dwork, F. McSherry, K. Nissim, and A. Smith, “Calibrating noise to sensitivity in private data analysis,” inTheory of cryptography conference. Springer, 2006, pp. 265–284

    work page 2006

  30. [32]

    The algorithmic foundations of differential privacy,

    C. Dwork, A. Rothet al., “The algorithmic foundations of differential privacy,”Foundations and Trends® in Theoretical Computer Science, vol. 9, no. 3–4, pp. 211–407, 2014

    work page 2014

  31. [33]

    R ´enyi differential privacy,

    I. Mironov, “R ´enyi differential privacy,” in2017 IEEE 30th Computer Security Foundations Symposium (CSF). IEEE, Aug. 2017, p. 263–275. [Online]. Available: http://dx.doi.org/10.1109/CSF.2017.11

    work page doi:10.1109/csf.2017.11 2017

  32. [34]

    Bandit-based communication-efficient client selection strategies for federated learn- ing,

    Y . Jee Cho, S. Gupta, G. Joshi, and O. Ya ˘gan, “Bandit-based communication-efficient client selection strategies for federated learn- ing,” in2020 54th Asilomar Conference on Signals, Systems, and Computers, 2020, pp. 1066–1069. 14

    work page 2020

  33. [35]

    Fedsampling: a better sampling strategy for federated learning,

    T. Qi, F. Wu, L. Lyu, Y . Huang, and X. Xie, “Fedsampling: a better sampling strategy for federated learning,” inProceedings of the Thirty- Second International Joint Conference on Artificial Intelligence, ser. IJCAI ’23, 2023

    work page 2023

  34. [36]

    Federated Learning with Non-IID Data

    Y . Zhao, M. Li, L. Lai, N. Suda, D. Civin, and V . Chandra, “Federated learning with non-iid data,”arXiv preprint arXiv:1806.00582, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  35. [37]

    Client selection in federated learning: Convergence analysis and power-of-choice selection strategies,

    Y . J. Cho, J. Wang, and G. Joshi, “Client selection in federated learning: Convergence analysis and power-of-choice selection strategies,” inarXiv preprint arXiv:2010.01243, 2020

    work page arXiv 2010

  36. [38]

    Communication-efficient federated learning via optimal client sampling,

    M. Ribero and H. Vikalo, “Communication-efficient federated learning via optimal client sampling,”arXiv preprint arXiv:2007.15197, 2020

    work page arXiv 2007

  37. [39]

    SCAFFOLD: Stochastic controlled averaging for federated learning,

    S. P. Karimireddy, S. Kale, M. Mohri, S. Reddi, S. Stich, and A. T. Suresh, “SCAFFOLD: Stochastic controlled averaging for federated learning,” inProceedings of the 37th International Conference on Machine Learning, ser. Proceedings of Machine Learning Research, H. D. III and A. Singh, Eds., vol. 119. PMLR, 13–18 Jul 2020, pp. 5132–5143

    work page 2020

  38. [40]

    Fairbatch: Batch selection for model fairness,

    Y . Roh, K. Lee, S. E. Whang, and C. Suh, “Fairbatch: Batch selection for model fairness,”arXiv preprint arXiv:2012.01696, 2020

    work page arXiv 2012

  39. [41]

    Data shapley: Equitable valuation of data for machine learning,

    A. Ghorbani and J. Zou, “Data shapley: Equitable valuation of data for machine learning,” inInternational conference on machine learning. PMLR, 2019, pp. 2242–2251

    work page 2019

  40. [42]

    Towards efficient data valuation based on the shapley value,

    R. Jia, D. Dao, B. Wang, F. A. Hubis, N. Hynes, N. M. G ¨urel, B. Li, C. Zhang, D. Song, and C. J. Spanos, “Towards efficient data valuation based on the shapley value,” inThe 22nd international conference on artificial intelligence and statistics. PMLR, 2019, pp. 1167–1176

    work page 2019

  41. [43]

    Beta shapley: a unified and noise-reduced data valuation framework for machine learning,

    Y . Kwon and J. Zou, “Beta shapley: a unified and noise-reduced data valuation framework for machine learning,”arXiv preprint arXiv:2110.14049, 2021

    work page arXiv 2021

  42. [44]

    Data valuation and detections in federated learning,

    W. Li, S. Fu, F. Zhang, and Y . Pang, “Data valuation and detections in federated learning,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2024, pp. 12 027–12 036

    work page 2024

  43. [45]

    Class-aware sample reweighting optimal transport for multi-source domain adaptation,

    S. Wang, B. Wang, Z. Zhang, A. A. Heidari, and H. Chen, “Class-aware sample reweighting optimal transport for multi-source domain adaptation,”Neurocomputing, vol. 523, pp. 213–223, 2023. [Online]. Available: https://www.sciencedirect.com/science/article/pii/ S0925231222015545

    work page 2023

  44. [46]

    Lava: Data valuation without pre-specified learning algorithms,

    H. A. Just, F. Kang, J. T. Wang, Y . Zeng, M. Ko, M. Jin, and R. Jia, “Lava: Data valuation without pre-specified learning algorithms,”arXiv preprint arXiv:2305.00054, 2023

    work page arXiv 2023

  45. [47]

    Gradient driven rewards to guarantee fairness in collaborative machine learning,

    X. Xu, L. Lyu, X. Ma, C. Miao, C. S. Foo, and B. K. H. Low, “Gradient driven rewards to guarantee fairness in collaborative machine learning,” inAdvances in Neural Information Processing Systems, M. Ranzato, A. Beygelzimer, Y . Dauphin, P. Liang, and J. W. Vaughan, Eds., vol. 34. Curran Associates, Inc., 2021, pp. 16 104– 16 117. [Online]. Available: http...

    work page 2021

  46. [48]

    Feature selection based on mu- tual information criteria of max-dependency, max-relevance, and min- redundancy,

    H. Peng, F. Long, and C. Ding, “Feature selection based on mu- tual information criteria of max-dependency, max-relevance, and min- redundancy,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 27, no. 8, pp. 1226–1238, 2005

    work page 2005

  47. [49]

    Differential evolution – a simple and efficient heuristic for global optimization over continuous spaces,

    R. Storn and K. Price, “Differential evolution – a simple and efficient heuristic for global optimization over continuous spaces,”J. of Global Optimization, vol. 11, no. 4, p. 341–359, Dec. 1997. [Online]. Available: https://doi.org/10.1023/A:1008202821328

    work page doi:10.1023/a:1008202821328 1997

  48. [50]

    Aarts and J

    E. Aarts and J. Korst,Simulated annealing and Boltzmann machines. John Wiley & Sons, 1989

    work page 1989

  49. [51]

    Stochastic relaxation, gibbs distributions, and the bayesian restoration of images,

    S. Geman and D. Geman, “Stochastic relaxation, gibbs distributions, and the bayesian restoration of images,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 6, no. 6, pp. 721–741, 1984

    work page 1984

  50. [52]

    Cooling schedules for optimal annealing,

    B. Hajek, “Cooling schedules for optimal annealing,”Mathematics of Operations Research, vol. 13, no. 2, pp. 311–329, 1988

    work page 1988

  51. [53]

    Optimization by simulated annealing,

    S. Kirkpatrick, C. D. Gelatt, and M. P. Vecchi, “Optimization by simulated annealing,”Science, vol. 220, no. 4598, pp. 671–680, 1983

    work page 1983

  52. [54]

    The theory and practice of simulated annealing,

    D. Henderson, S. H. Jacobson, and A. W. Johnson, “The theory and practice of simulated annealing,” inHandbook of metaheuristics. Springer, 2003, pp. 287–319

    work page 2003

  53. [55]

    Deep learning with differential privacy,

    M. Abadi, A. Chu, I. Goodfellow, H. B. McMahan, I. Mironov, K. Talwar, and L. Zhang, “Deep learning with differential privacy,” in Proceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security, ser. CCS’16. ACM, Oct. 2016

    work page 2016

  54. [56]

    Resolving individuals contributing trace amounts of dna to highly complex mix- tures using high-density snp genotyping microarrays,

    N. Homer, S. Szelinger, M. Redman, D. Duggan, W. Tembe, J. Muehling, J. Pearson, D. Stephan, S. Nelson, and D. Craig, “Resolving individuals contributing trace amounts of dna to highly complex mix- tures using high-density snp genotyping microarrays,”PLoS genetics, vol. 4, p. e1000167, 09 2008

    work page 2008

  55. [57]

    Differential privacy and the risk-utility tradeoff for multi-dimensional contingency tables,

    S. E. Fienberg, A. Rinaldo, and X. Yang, “Differential privacy and the risk-utility tradeoff for multi-dimensional contingency tables,” inPrivacy in Statistical Databases, J. Domingo-Ferrer and E. Magkos, Eds. Berlin, Heidelberg: Springer Berlin Heidelberg, 2010, pp. 187–199

    work page 2010

  56. [58]

    Structure and sensitivity in differential privacy: Comparing k-norm mechanisms,

    J. Awan and A. Slavkovi ´c, “Structure and sensitivity in differential privacy: Comparing k-norm mechanisms,”Journal of the American Statistical Association, vol. 116, no. 534, pp. 935–954, 2021

    work page 2021

  57. [59]

    Parallel simulated annealing techniques,

    D. R. Greening, “Parallel simulated annealing techniques,”Physica D: Nonlinear Phenomena, vol. 42, no. 1, pp. 293–306, 1990. [Online]. Available: https://www.sciencedirect.com/science/article/pii/ 0167278990900843

    work page 1990

  58. [60]

    Clusterfl: a similarity-aware federated learning system for human activity recogni- tion,

    X. Ouyang, Z. Xie, J. Zhou, J. Huang, and G. Xing, “Clusterfl: a similarity-aware federated learning system for human activity recogni- tion,” inProceedings of the 19th Annual International Conference on Mobile Systems, Applications, and Services, ser. MobiSys ’21. New York, NY , USA: Association for Computing Machinery, 2021, p. 54–66

    work page 2021

  59. [61]

    On the problem of the most efficient tests of statistical hypotheses,

    J. Neyman and E. S. Pearson, “On the problem of the most efficient tests of statistical hypotheses,”Philosophical Transactions of the Royal Society of London. Series A, Containing Papers of a Mathematical or Physical Character, vol. 231, pp. 289–337, 1933. [Online]. Available: http://www.jstor.org/stable/91247 APPENDIXA GROUP FAIRNESS METRICS Group fairne...

    work page 1933