Pith. sign in

REVIEW 4 major objections 6 minor 46 references

FedCFA: Alleviating Simpson's Paradox in Model Aggregation with Counterfactual Federated Learning

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read FedCFA uses counterfactual generation to align local data with the global distribution, mitigating Simpson's Paradox in federated learning.

desk verdict A plausible empirical method for heterogeneous FL with a broken theoretical justification; worth refereeing, but only with code and a rewritten proof. read the letter →

arxiv 2412.18904 v1 pith:LHBJ2UGT submitted 2024-12-25 cs.LG

Zhonghua Jiang , Jimin Xu , Shengyu Zhang , Tao Shen , Jiwei Li , Kun Kuang , Haibin Cai , Fei Wu This is my paper · ORCID
classification cs.LG
keywords FederatedlearningSimpson'sparadoxCounterfactualDataheterogeneityNon-IIDFactordecorrelationContrastiveModelaggregation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that federated learning's standard averaging can fall victim to Simpson's Paradox: a trend that holds in each client's local data disappears or reverses in the aggregated global model, so the global model encodes wrong feature-label associations. The authors propose FedCFA, which builds a small global-average dataset on the server, sends it to clients, and uses it to generate counterfactual positive and negative samples for each local batch. Replacing a sample's low- or high-gradient latent factors with the corresponding factors from global-average data pulls each client's training distribution toward the global one and teaches the model to ignore spurious associations. If the claim is right, the same framework that fixes Simpson's Paradox also improves accuracy and convergence speed under both Non-IID and IID partitions, on image and text datasets.

What carries the argument

The central mechanism is the global average dataset D̄g = {X̄g, Ȳg}, built by each client averaging B random subsets of its local data and the server aggregating those averages; this dataset serves as the reference distribution for global data. A counterfactual transformation module extracts latent factors with an encoder, selects the top-k low- or high-gradient factors using decoder gradients, and replaces them with the corresponding factors from the global average data, producing positive samples that keep the original label and negative samples that receive a weighted counterfactual label. A factor decorrelation (FDC) loss penalizes the mean absolute Pearson correlation coefficient between factor variables, forcing the encoder to produce more independent factors so that the counterfactual samples cleanly isolate the features being manipulated.

What would settle it

Compute the Wasserstein distance W(P_{F̄g}, P_{F'_k}) and W(P_{F̄g}, P_{F_k}) with the trained encoder on a natural-image dataset; if the post-counterfactual distance is not smaller, or if human raters cannot recognize the class of the generated counterfactual images, the claimed distribution-alignment mechanism does not hold.

Watch

Extended reading notes

Core claim

The paper's central claim is that Simpson's Paradox in federated aggregation can be mitigated by counterfactual data generation guided by global average data, rather than by aligning client models to a global model that may itself be untrustworthy. FedCFA replaces critical local latent factors with factors extracted from a global average dataset, producing positive and negative counterfactual samples that make each client's local distribution closer to the global distribution. The authors further claim that adding a factor decorrelation loss makes the extracted factors more independent and improves counterfactual sample quality. Experiments on six datasets report that FedCFA achieves higher global model accuracy than six baselines after 500 communication rounds and reaches target accuracy in fewer rounds.

Load-bearing premise

The load-bearing premise is that a pixel-wise average of each client's data subsets, pooled across clients, faithfully represents the global data distribution, and that replacing factors with features from that average yields valid, class-preserving counterfactual samples.

Editorial extensions

If this is right

  • On CIFAR10 and CIFAR100 under six partitions, FedCFA reports the highest top-1 accuracy after 500 rounds, with the largest reported margin being 7.75 percentage points over the best baseline on CIFAR100 with Dir100(0.6).
  • In CIFAR10 target-accuracy tests, FedCFA reaches the targets in 375 to 453 rounds while most baselines fail to reach them within 1000 rounds, implying lower communication cost.
  • Ablation studies show that both counterfactual modules contribute to the accuracy gain and that their combined effect exceeds the sum of the individual effects, and adding the FDC loss further improves accuracy.
  • FedCFA maintains its advantage when the number of clients grows from 60 to 100 and also under IID partitions, suggesting the method is not limited to strongly heterogeneous label distributions.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The Wasserstein-distance proof assumes that every replaced factor moves closer to the corresponding global-average factor; in practice, on natural images, some replacements may land off the data manifold, so the observed gains could partly come from a regularizing effect rather than from pure distribution alignment.
  • Because the global average dataset is shared with all clients, it communicates aggregate statistics about the training data; a privacy-focused extension would need to perturb or distill this shared reference before sending it to clients.
  • A testable extension is to construct the global reference in latent space or from per-class prototypes instead of pixel-space averages; that would likely improve counterfactual quality on datasets with large intra-class variation, where pixel averages quickly become unrecognizable.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper proposes FedCFA, a federated learning framework that aims to mitigate Simpson's Paradox in model aggregation. The method constructs a global average dataset by averaging raw inputs (and labels) across clients, extracts factor representations with an encoder, and generates counterfactual positive/negative samples by replacing selected local factors with features of the global average data. A factor decorrelation (FDC) loss based on Pearson correlation is added to encourage factor independence. The authors claim that these counterfactual samples align local distributions with the global distribution and show improved accuracy and convergence over six baselines on six datasets under IID and Non-IID partitions. The paper includes a proof attempt via Wasserstein distance, extensive tables of classification accuracy, ablations of the counterfactual modules, and a synthetic Simpson's Paradox dataset experiment.

Significance. If correct, FedCFA would be a useful new approach to a real and under-studied failure mode in federated learning: Simpson's Paradox where a global model aggregated from client models contradicts the true global distribution. The paper gives a concrete algorithmic recipe, compares against six baselines on six datasets, provides ablations of each loss component, and constructs a dedicated Simpson's Paradox benchmark. These are valuable contributions. However, the central distribution-alignment claim is supported by a flawed proof and by experiments whose reproducibility is limited by unreleased code, single-run rounds-to-target results, and hyperparameter selection on the same datasets. The empirical improvements are plausible but not yet established at the standard needed for publication.

major comments (4)
  1. [Proof (Eqs. 13-17)] The Wasserstein proof is invalid. Eq. (16) does not follow from the construction in Eqs. (6)-(7): only the selected factors are replaced, so for any j not in the key-factor set, ||fg - f'_k,j|| = ||fg - f_k,j|| and the strict inequality is false. Even for replaced factors, the equality f'_k,j = fbar_g,j holds only for the specific global-average factor used, not for every element of Fbar_g. Moreover, the step from a pointwise norm inequality to an inequality between infima over different couplings in Eq. (17) is not justified: the optimal transport couplings for (P_Fbar_g, P_F'_k) and (P_Fbar_g, P_F_k) are generally different, so a pointwise bound cannot be transferred to the infimum. Since this proof is the only theoretical support for the paper's central claim that counterfactual transformation brings local data closer to the global distribution, it must be repaired or replaced by direct empirical distribution-distance measurements (e.g., MMD or Wasserstein estimates before and after counterfactual transformation).
  2. [Global Average Dataset Construction (Eqs. 3-4)] The method assumes that a pixel-space average of raw inputs and a corresponding average of labels faithfully approximate the global data distribution. For natural images, the mean of a set of images is typically off the data manifold and can be uninterpretable, and averaging one-hot labels produces soft label vectors that do not correspond to any actual class. The counterfactual samples in Eqs. (6)-(8) inherit these properties, so the claim that Lpos and Lneg align local distributions with the global distribution rests on an unverified premise. The paper should provide evidence that the constructed global average dataset is a faithful proxy, for example by visualizing generated counterfactual samples, measuring distribution distance before/after transformation, and ablating the average-dataset size B and subset size n.
  3. [Experiments: hyperparameter selection and validation protocol] Several hyperparameters (lambda_pos, lambda_neg, lambda_corr, topk, hook, and the global average dataset size B) are tuned on the same datasets for which the final results are reported, with best results selected. Figure 6 and Table 5 show that accuracy varies by several points with these settings; for instance, Table 5 reports an 80.01 accuracy at topk=24 but 78.53 at topk=8, and hook=2 gives 80.01 while hook=4 gives 75.31. Without a fixed train/validation/test split for hyperparameter selection or a nested protocol, the margins in Tables 1-3 may reflect selection bias. The authors should specify the exact protocol, report the chosen configuration, and give results on held-out test partitions across multiple seeds.
  4. [Experiments: reproducibility and statistical reporting] No code or data are released, and Table 3 reports rounds-to-target without error bars or repetition counts. The empirical claim is the main support for the method, so the paper should include code or detailed pseudocode, dataset partition metadata, the number of random seeds, and confidence intervals for all headline numbers, including Table 3 and Figure 4. Table 3 should also clarify how the target accuracies were set before running baselines, since targets appear close to FedCFA's achieved values.
minor comments (6)
  1. [Eq. (8)] The construction of Yneg as a weighted average of the local label and the global average label is not explained; if Ybar_g is a soft label derived from averaged one-hot vectors, the resulting Yneg is a convex combination that may not correspond to any valid class. Please clarify how the classification loss Lneg handles these soft labels.
  2. [Baseline list and Table 1] The baseline is called FedPVR in Section 'Baseline Methods' and FedPRV in Table 1; please make the naming consistent.
  3. [Figure 4] The Simpson's Paradox experiment reports only a learning curve without final accuracy values, error bars, or a comparison with FedProx/SCAFFOLD; quantitative results and repetitions should be provided.
  4. [Algorithm 1, lines 13-16] The server receives local average datasets (Xbar_k, Ybar_k) from clients and broadcasts the global average dataset. The privacy implications of sharing data-derived averages are not discussed; in a federated setting this requires at least a brief treatment.
  5. [Eq. (10)] The summation index |F0| appears in the denominator and range of Eq. (10), while the factor-set size is elsewhere denoted |F| or |F0|; please standardize the notation.
  6. [Proof, sentence before Eq. (15)] The reduction from W(P_Fg, .) to W(P_Fbar_g, .) via 'Since we construct a global average dataset to approximate the global data distribution' is asserted without a quantitative approximation bound; if this step is retained, it needs a precise statement.

Circularity Check

1 steps flagged · score 6.0 of 10

The theoretical proof of distribution alignment reduces to the counterfactual replacement construction, while the experimental comparisons remain independent.

  1. self definitional [Method, Eqs. 6–7; Proof, Eqs. 16–17]
    "Fpos = M askpos ∗ F + (1 − M askpos) ∗ ¯Fg, (6) Fneg = M askneg ∗ F + (1 − M askneg) ∗ ¯Fg, (7) ... The counterfactual transformation generates F′k by replacing the key factors Sk of the local factor Fk with the corresponding factors in global average data factor ¯Fg. This means that for any fk,j ∈ Fk, its transformed sample f′k,j has a higher similarity with ¯fg,j ... Therefore, for any ¯fg ∈ ¯Fg and f′k,j ∈ F′k, we have ||¯fg − f′k,j|| < ||¯fg − fk,j||. From this, we can conclude ... W1 < W0."

    Eqs. 6–7 define the counterfactual feature vectors by literally copying selected factor coordinates from the global-average feature vector ¯Fg. The proof's Eq. 16 is not derived; it is asserted as 'this means' the replacement made each transformed sample closer to ¯Fg. Eq. 17 then converts that pointwise assertion into a Wasserstein inequality without analyzing the infimum over couplings. Thus the claimed theorem that the counterfactual local distribution is closer to the global distribution is simply the intended effect of the replacement construction restated, not an independent first-principles result. The alignment claim therefore reduces, in the proof, to the definition of the counterfactual samples.

full rationale

FedCFA's empirical evaluation is self-contained: reported accuracy and rounds-to-target are measured against external baselines on standard benchmarks, and hyperparameters are tuned or ablated rather than fitted to the reported targets. There is no load-bearing self-citation: the cited counterfactual-learning works are related-work context, not the basis of the framework. The single circular element is the theoretical 'Proof' of distribution alignment. Eqs. 6–7 construct counterfactual features by copying selected coordinates from the global-average feature vector; Eq. 16 restates this construction as pointwise closeness, and Eq. 17 asserts, without the required optimal-transport argument, that Wasserstein distance to the global-average factor distribution must decrease. Pointwise closeness under a fixed construction-specific pairing does not imply a smaller infimum over all couplings, so the conclusion is not entailed by the premises; the proof simply labels the construction's intended effect as a theorem. Since this proof is the paper's only theoretical support for the central 'aligning local distributions with the global' claim, the central claim is partially circular; the experimental gains remain independent evidence.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central mechanism relies on several unproven or domain-specific assumptions: the global average dataset approximates the global distribution, latent factor replacement produces valid counterfactual samples, and the pointwise distance inequality underlying the Wasserstein proof. These are not derived from first principles and are not independently verified. Hyperparameters are tuned on the evaluation datasets, adding further degrees of freedom.

free parameters (6)
  • lambda_pos = tuned, e.g., 5 (best in grid)
    Weight for positive counterfactual loss; grid searched over {0.1,0.5,1,2,5,10} in Figure 6.
  • lambda_neg = tuned, e.g., 5 (best in grid)
    Weight for negative counterfactual loss; grid searched similarly.
  • lambda_corr = 0.1 (best)
    Weight for FDC loss; grid searched over {0.001,0.01,0.1,1.0} in Table IV.
  • topk = 24 (best)
    Number of factors replaced in counterfactual transformation; grid over {8,16,24,32} in Table 5.
  • hook = 2 (best)
    Layer index in ResNet18 where counterfactual transformation is applied; grid over {0,1,2,4} in Table 5.
  • B (global average dataset size) = not reported
    Size of the global average dataset; algorithm references B but no value is given.
assumptions (4)
  • standard math Central Limit Theorem justifies using averaged subsets as estimates of the global mean.
    Invoked in Global Average Dataset Construction (Eq. 3), but applied to arbitrary data samples, not just i.i.d. scalar observations.
  • domain assumption The global average dataset approximates the global data distribution.
    The whole method relies on this; no evidence that pixel-space averages are representative.
  • domain assumption Replacing latent factors with global average features yields valid, class-preserving counterfactual samples.
    Core to counterfactual transformation; FDC loss only enforces linear decorrelation and cannot guarantee semantic disentanglement.
  • ad hoc to paper For any pair of samples, the transformed feature is closer to the global average feature than the original (Eq 16).
    This unproven pointwise inequality is the basis of the Wasserstein proof.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FedCFA: Alleviating Simpson's Paradox in Model Aggregation with Counterfactual Federated Learning." pith.science (2026). https://pith.science/paper/LHBJ2UGT

@misc{pith2026241218904,
  author       = {Pith},
  title        = {Pith review of: FedCFA: Alleviating Simpson's Paradox in Model Aggregation with Counterfactual Federated Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LHBJ2UGT}},
  note         = {Machine review of arXiv:2412.18904}
}
read the original abstract

Federated learning (FL) is a promising technology for data privacy and distributed optimization, but it suffers from data imbalance and heterogeneity among clients. Existing FL methods try to solve the problems by aligning client with server model or by correcting client model with control variables. These methods excel on IID and general Non-IID data but perform mediocrely in Simpson's Paradox scenarios. Simpson's Paradox refers to the phenomenon that the trend observed on the global dataset disappears or reverses on a subset, which may lead to the fact that global model obtained through aggregation in FL does not accurately reflect the distribution of global data. Thus, we propose FedCFA, a novel FL framework employing counterfactual learning to generate counterfactual samples by replacing local data critical factors with global average data, aligning local data distributions with the global and mitigating Simpson's Paradox effects. In addition, to improve the quality of counterfactual samples, we introduce factor decorrelation (FDC) loss to reduce the correlation among features and thus improve the independence of extracted factors. We conduct extensive experiments on six datasets and verify that our method outperforms other FL methods in terms of efficiency and global model accuracy under limited communication rounds.

Figures

Figures reproduced from arXiv: 2412.18904 by the authors.

Figure 1
Figure 1. Simpson’s Paradox. Client 𝑖 local dataset 𝐷! White Cat White Cat Black Cat Red Cat White Dog Black Dog Black Dog White Dog Yellow Dog Black Cat White ‘cat’ Black ‘dog’ Cat ‘cat’ Dog ‘dog’ Counterfactual positive/negative samples FedCFA [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. FedCFA can generate counterfactual samples that [PITH_FULL_IMAGE:figures/full_fig_p001_2.png] view at source ↗
Figure 3
Figure 3. Local model training process in FedCFA. Based on the above analysis, we construct a global av￾erage dataset of size B to approximate global data dis￾tribution. Specifically, each client randomly divides lo￾cal data Dk into B subsets of n = ⌊|Dk|/B⌋ samples: {x(i−1)n+1, x(i−1)n+2, . . . , xin}, i = 1, 2, . . . , B. For each subset, calculate its mean x¯i : x¯i = 1 n Pi·n j=(i−1)n+1 xj . (4) As a result, the client ca… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: The top-1 accuracy (%) on MNIST data with [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Error Bar Chart: At experimental settings of dif [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: The top-1 accuracy (%) of FedCFA with different [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 34 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    L.; Zippo, A

    Alfeo, A. L.; Zippo, A. G.; Catrambone, V.; Cimino, M. G.; Toschi, N.; and Valenza, G. 2023. From local counterfactuals to global feature importance: efficient, robust, and model-agnostic explanations for brain connectivity networks. Computer Methods and Programs in Biomedicine, 236: 107550

    2023

  4. [4]

    Caldas, S.; Duddu, S. M. K.; Wu, P.; Li, T.; Kone c n \`y , J.; McMahan, H. B.; Smith, V.; and Talwalkar, A. 2018. Leaf: A benchmark for federated settings. arXiv preprint arXiv:1812.01097

    arXiv 2018

  5. [5]

    Cao, X.; and Gong, N. Z. 2022. Mpaf: Model poisoning attacks to federated learning based on fake clients. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 3396--3404

    2022

  6. [6]

    Chai, B.; Liu, K.; and Yang, R. 2022. Cross-Domain Federated Data Modeling on Non-IID Data. Computational Intelligence and Neuroscience, 2022

    work page 2022

  7. [7]

    Chen, H.; and Vikalo, H. 2023. Federated learning in non-iid settings aided by differentially private synthetic data. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 5026--5035

    work page 2023

  8. [8]

    S.; and Cheng, J

    Cheng, A.; Wang, P.; Zhang, X. S.; and Cheng, J. 2022. Differentially private federated learning with local regularization and sparsification. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 10122--10131

    work page 2022

Show all 46 references
  1. [9]

    Chow, K.-H.; Liu, L.; Wei, W.; Ilhan, F.; and Wu, Y. 2023. STDLens: Model Hijacking-resilient Federated Learning for Object Detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 16343--16351

    2023

  2. [10]

    Fang, X.; and Ye, M. 2022. Robust federated learning with noisy and heterogeneous clients. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 10072--10081

    2022

  3. [11]

    Gao, L.; Fu, H.; Li, L.; Chen, Y.; Xu, M.; and Xu, C.-Z. 2022. Feddc: Federated learning with non-iid data via local drift decoupling and correction. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 10112--10121

    2022

  4. [12]

    Go, A.; Bhayani, R.; and Huang, L. 2009. Twitter sentiment classification using distant supervision. CS224N project report, Stanford, 1(12): 2009

    2009

  5. [13]

    Hamman, F.; and Dutta, S. 2024. Demystifying Local & Global Fairness Trade-offs in Federated Learning Using Partial Information Decomposition. In The Twelfth International Conference on Learning Representations

    2024

  6. [14]

    J.; Chen, C.; and Duke, L

    Hao, W.; El-Khamy, M.; Lee, J.; Zhang, J.; Liang, K. J.; Chen, C.; and Duke, L. C. 2021. Towards fair federated learning with zero-shot data augmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 3310--3319

    2021

  7. [15]

    Ilhan, F.; Su, G.; and Liu, L. 2023. ScaleFL: Resource-Adaptive Federated Learning With Heterogeneous Clients. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 24532--24541

    2023

  8. [16]

    P.; Kale, S.; Mohri, M.; Reddi, S.; Stich, S.; and Suresh, A

    Karimireddy, S. P.; Kale, S.; Mohri, M.; Reddi, S.; Stich, S.; and Suresh, A. T. 2020. Scaffold: Stochastic controlled averaging for federated learning. In International conference on machine learning, 5132--5143. PMLR

    2020

  9. [17]

    B.; Yu, F

    Kone c n \`y , J.; McMahan, H. B.; Yu, F. X.; Richt \'a rik, P.; Suresh, A. T.; and Bacon, D. 2016. Federated learning: Strategies for improving communication efficiency. arXiv preprint arXiv:1610.05492

    2016 arXiv

  10. [18]

    Krizhevsky, A.; Hinton, G.; et al. 2009. Learning multiple layers of features from tiny images

    2009

  11. [19]

    Li, A.; Sun, J.; Zeng, X.; Zhang, M.; Li, H.; and Chen, Y. 2021 a . Fedmask: Joint computation and communication-efficient personalized federated learning via heterogeneous masking. In Proceedings of the 19th ACM Conference on Embedded Networked Sensor Systems, 42--55

    2021

  12. [20]

    N.; Alstr m, T

    Li, B.; Esfandiari, Y.; Schmidt, M. N.; Alstr m, T. S.; and Stich, S. U. 2023 a . Synthetic data shuffling accelerates the convergence of federated learning under data heterogeneity. arXiv preprint arXiv:2306.13263

    2023 arXiv

  13. [21]

    N.; Alstr m, T

    Li, B.; Schmidt, M. N.; Alstr m, T. S.; and Stich, S. U. 2023 b . On the effectiveness of partial variance reduction in federated learning with heterogeneous data. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 3964--3973

    2023

  14. [22]

    Li, Q.; He, B.; and Song, D. 2021. Model-contrastive federated learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 10713--10722

    2021

  15. [23]

    Li, T.; Hu, S.; Beirami, A.; and Smith, V. 2021 b . Ditto: Fair and robust federated learning through personalization. In International Conference on Machine Learning, 6357--6368. PMLR

    2021

  16. [24]

    K.; Talwalkar, A.; and Smith, V

    Li, T.; Sahu, A. K.; Talwalkar, A.; and Smith, V. 2020 a . Federated learning: Challenges, methods, and future directions. IEEE signal processing magazine, 37(3): 50--60

    2020

  17. [25]

    K.; Zaheer, M.; Sanjabi, M.; Talwalkar, A.; and Smith, V

    Li, T.; Sahu, A. K.; Zaheer, M.; Sanjabi, M.; Talwalkar, A.; and Smith, V. 2020 b . Federated optimization in heterogeneous networks. Proceedings of Machine learning and systems, 2: 429--450

    2020

  18. [26]

    Li, T.; Sanjabi, M.; Beirami, A.; and Smith, V. 2019. Fair resource allocation in federated learning. arXiv preprint arXiv:1905.10497

    2019 arXiv

  19. [27]

    Li, Z.; Zhang, J.; Liu, L.; and Liu, J. 2022. Auditing privacy defenses in federated learning via generative gradient leakage. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 10132--10142

    2022

  20. [28]

    Liao, D.; Gao, X.; Zhao, Y.; and Xu, C.-Z. 2023. Adaptive Channel Sparsity for Federated Learning Under System Heterogeneity. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 20432--20441

    2023

  21. [29]

    Luo, K.; Li, X.; Lan, Y.; and Gao, M. 2023. GradMA: A Gradient-Memory-based Accelerated Federated Learning with Alleviated Catastrophic Forgetting. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 3708--3717

    2023

  22. [30]

    McMahan, B.; Moore, E.; Ramage, D.; Hampson, S.; and y Arcas, B. A. 2017. Communication-efficient learning of deep networks from decentralized data. In Artificial intelligence and statistics, 1273--1282. PMLR

    2017

  23. [31]

    Mendieta, M.; Yang, T.; Wang, P.; Lee, M.; Ding, Z.; and Chen, C. 2022. Local learning matters: Rethinking data heterogeneity in federated learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 8397--8406

    2022

  24. [32]

    Mou, Y.; Geng, J.; Welten, S.; Rong, C.; Decker, S.; and Beyan, O. 2021. Optimized federated learning on class-biased distributed data sources. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases, 146--158. Springer

    2021

  25. [33]

    R.; Dey, E.; Roy, N.; and Gangopadhyay, A

    Ovi, P. R.; Dey, E.; Roy, N.; and Gangopadhyay, A. 2023. Mixed Quantization Enabled Federated Learning to Tackle Gradient Inversion Attacks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 5045--5053

    2023

  26. [34]

    P.; Xia, Y.; Wang, F.; Adeli, E.; Fei-Fei, L.; and Rubin, D

    Qu, L.; Zhou, Y.; Liang, P. P.; Xia, Y.; Wang, F.; Adeli, E.; Fei-Fei, L.; and Rubin, D. 2022. Rethinking architecture design for tackling data heterogeneity in federated learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 10061--10071

    2022

  27. [35]

    E.; Kamili, M

    Reguieg, H.; Hanjri, M. E.; Kamili, M. E.; and Kobbane, A. 2023. A Comparative Evaluation of FedAvg and Per-FedAvg Algorithms for Dirichlet Distributed Heterogeneous Data. arXiv preprint arXiv:2309.01275

    2023 arXiv

  28. [36]

    Rothchild, D.; Panda, A.; Ullah, E.; Ivkin, N.; Stoica, I.; Braverman, V.; Gonzalez, J.; and Arora, R. 2020. Fetchsgd: Communication-efficient federated learning with sketching. In International Conference on Machine Learning, 8253--8265. PMLR

    2020

  29. [37]

    Seo, E.; and Elmroth, E. 2023. MadFed: Enhancing Federated Learning with Marginal-data Model Fusion. IEEE Access

    2023

  30. [38]

    Xie, Z.; and Song, S. 2023. Fedkl: Tackling data heterogeneity in federated reinforcement learning by penalizing kl divergence. IEEE Journal on Selected Areas in Communications, 41(4): 1227--1242

    2023

  31. [39]

    J.; and Yang, E

    Yoon, T.; Shin, S.; Hwang, S. J.; and Yang, E. 2021. Fedmix: Approximation of mixup under mean augmented federated learning. arXiv preprint arXiv:2107.00233

    2021 arXiv

  32. [40]

    Zeng, D.; Liang, S.; Hu, X.; Wang, H.; and Xu, Z. 2023. FedLab: A Flexible Federated Learning Framework. Journal of Machine Learning Research, 24(100): 1--7

    2023

  33. [41]

    Zhang, C.; Xie, Y.; Bai, H.; Yu, B.; Li, W.; and Gao, Y. 2021. A survey on federated learning. Knowledge-Based Systems, 216: 106775

    2021

  34. [42]

    Zhang, S.; Jiang, T.; Wang, T.; Kuang, K.; Zhao, Z.; Zhu, J.; Yu, J.; Yang, H.; and Wu, F. 2020. Devlbert: Learning deconfounded visio-linguistic representations. In Proceedings of the 28th ACM International Conference on Multimedia, 4373--4382

    2020

  35. [43]

    Zhang, S.; Miao, Q.; Nie, P.; Li, M.; Chen, Z.; Feng, F.; Kuang, K.; and Wu, F. 2024. Transferring Causal Mechanism over Meta-representations for Target-Unknown Cross-domain Recommendation. ACM Transactions on Information Systems, 42(4): 1--27

    2024

  36. [44]

    C.; Elkordy, A

    Zhao, J. C.; Elkordy, A. R.; Sharma, A.; Ezzeldin, Y. H.; Avestimehr, S.; and Bagchi, S. 2023. The Resource Problem of Using Linear Layer Leakage Attack in Federated Learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 3974--3983

    2023

  37. [45]

    Zhu, H.; Xu, J.; Liu, S.; and Jin, Y. 2021. Federated learning on non-IID data: A survey. Neurocomputing, 465: 371--390

    2021

  38. [46]

    Zhu, J.; Ma, X.; and Blaschko, M. B. 2023. Confidence-aware personalized federated learning via variational expectation maximization. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 24542--24551

    2023

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.