Pith. sign in

REVIEW 4 major objections 5 minor 66 references

Harmonizing Generalization and Personalization in Ring-topology Decentralized Federated Learning

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read A ring-topology federated learning method that splits each client model into a personalized PersonaNet and a shared Learngene improves both local and global accuracy under data heterogeneity while exchanging only 0.58 MB of parameters per…

desk verdict A communication-efficient personalized DFL method with consistent empirical gains, but the 'global' class statistics are mostly local under the stated hyperparameters and the convergence proof is hand-wavy; worth a serious but critical referee. read the letter →

arxiv 2504.19103 v1 pith:CSR4QKUY submitted 2025-04-27 cs.LG cs.DC

Shunxin Guo , Jiaqi Lv , Xin Geng This is my paper · ORCID
classification cs.LGcs.DC
keywords decentralizedfederatedlearningringtopologypersonalizationgeneralizationdataheterogeneityGaussianmixtureadversarialLearngene
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 the conflict between generalization and personalization in serverless ring-topology federated learning can be resolved by splitting each client's model into two modules: a private PersonaNet that models class-specific features as a Gaussian mixture, and a shared Learngene that learns class-invariant representations through adversarial training. Only the Learngene and per-class mean/covariance statistics travel along the ring, so each round exchanges about 0.58 MB of parameters rather than full models. The claim is that this division lets clients keep their own data distributions while still converging to a common, transferable representation, improving both local test accuracy and global test accuracy across Dirichlet and shard-based non-IID settings. If true, it would give decentralized federated learning a way to get the benefits of shared knowledge and communication efficiency without a central server.

What carries the argument

The load-bearing mechanism is the divide-and-conquer latent decomposition with ring-based averaging. Each client's model is written as $w_m = [\psi_m, \phi, \theta_m, \omega_m]$, where $\psi_m$ is PersonaNet, $\phi$ is the shared Learngene, $\theta_m$ the decoder, and $\omega_m$ the classifier. On receiving its neighbor's Learngene and class statistics, the client sets $\phi \leftarrow (\tilde{\phi}+\phi)/2$ and updates $\mu_k, \Sigma_k$ by exponential moving average with $\alpha=0.99$, then trains locally with the ELBO objective: reconstruction, KL divergence to the class-conditional Gaussian prior for PersonaNet, and KL divergence plus an adversarial uniform-class loss for Learngene. The adversarial classifier, with parameters $\vartheta$, and the uniform-prior term enforce that $z_l$ carries no class identity, which is what makes the module transferable around the ring.

What would settle it

Track the per-hop change in the shared knowledge encoder, $\|\tilde{\phi}^t - \phi_m^t\|_2^2$, and the Local-T/Global-T accuracies while increasing label skew, for example moving from Dir(0.4) to Dir(0.05) or to shards with s=2; if the drift grows sharply and accuracy collapses when two clients with disjoint label sets pass the model, the paper's convergence and balance claims would fail.

Watch

Extended reading notes

Core claim

DRDFL's central discovery is that personalization and generalization need not be traded off if they are assigned to separate latent channels. PersonaNet maps each input to a class-conditioned latent $z_p$ regularized toward a per-class Gaussian prior with mean $\mu_k$ and covariance $\Sigma_k$ updated by exponential moving average with the inherited global statistics, so the representation stays discriminative for the client's own label distribution. Learngene maps the same input to a class-independent latent $z_l$, regularized toward $\mathcal{N}(0,I)$ and pushed by an adversarial classifier toward a uniform class distribution, so the representation captures knowledge that transfers across clients. The two latents are concatenated and fed to a decoder; reconstruction with injected noise and a classifier on both original and perturbed data completes the objective. The claim is that this two-channel disentanglement lets the ring propagate a stable consensus representation while each client retains its personalized head, and the experiments report gains up to 3.28% in local test accuracy over eight comparison methods while sharing only the 0.58 MB Learngene and class statistics per iteration.

Load-bearing premise

The convergence proof assumes that the shared knowledge encoder barely changes when one client's copy is averaged with the next client's copy along the ring; whether that drift stays small under extreme label skew is what determines whether the balance holds.

Editorial extensions

If this is right

  • Clients can keep the personalized part of their model private while sharing only the Learngene and class-level Gaussian statistics, cutting per-round communication from hundreds of millions of parameters to 0.58 MB.
  • The same architecture produces high Local-T and high Global-T under both Dirichlet (β=0.1, 0.4) and shard-based (s=4,5,20,30) non-IID splits, whereas personalized baselines typically sacrifice one for the other.
  • New clients joining the ring after 200 rounds of training converge faster when initialized with the inherited Learngene and global priors than with a sparse-mask initializer.
  • Because raw gradients and instance-level representations are never exchanged, gradient-inversion attacks cannot reconstruct recognizable training images from the transmitted information.
  • Under the stated assumptions, each client's non-convex objective converges at rate $O(1/T)$, so the dual-module split does not by itself introduce a convergence barrier.

Reading between the lines

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

  • The same two-channel split could be lifted out of the ring: the Learngene could serve as a task-agnostic feature extractor for newly arriving domains or tasks, since its adversarial uniform-class training makes the shared latent class-independent; the paper only demonstrates fast convergence for new clients with unseen distributions, not transfer to new class sets.
  • A natural stress test would vary the EMA smoothing factor α rather than fixing it at 0.99; under fast-moving class distributions, a fixed slow EMA may lag, and adapting α per class could extend DRDFL to non-stationary federated settings without changing the architecture.
  • The privacy claim rests on the assumption that per-class means and covariances leak nothing about individual instances; that is plausible but unproven, so a membership-inference test on the shared Gaussian statistics would tell whether the privacy benefit is as strong as the gradient-inversion reconstruction experiment suggests.
  • If the learned shared representation is truly class-invariant, it could double as a calibration or out-of-distribution detector; a reader could test by checking whether the shared latent separates known from novel classes, which the paper does not do.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes DRDFL, a ring-topology decentralized federated learning method that splits each client model into a personalized PersonaNet and a shared Learngene module. PersonaNet models class-conditional Gaussian latent distributions with EMA-updated class statistics, while Learngene is trained with a KL prior and an adversarial classifier to extract class-invariant representations. The two modules are combined in a VAE-style decoder, and only the Learngene plus class statistics are exchanged along the ring. Experiments on SVHN, CIFAR-10, and CIFAR-100 under Dirichlet and shard-based non-IID partitions compare DRDFL with eight baselines, reporting higher Local-T and Global-T accuracies with only 0.58 MB of communication per round. The paper also provides an ablation study, convergence curves, Grad-CAM visualizations, a new-client adaptation experiment, and a gradient-inversion robustness study, plus a convergence proof in Appendix D.

Significance. If the claims hold, DRDFL is a useful contribution to decentralized FL: it demonstrates a parameter-efficient way to combine personalization and generalization under ring-topology constraints, and the additional experiments on new-client adaptation and resistance to gradient inversion are valuable. The strongest assets are the low communication cost (0.58M parameters exchanged), the consistent gains across three datasets and multiple heterogeneity settings, and the explicit ablation showing both loss components matter. However, the significance is tempered by three correctness gaps: the convergence proof relies on an unproved approximation, the adversarial objective appears internally inconsistent as written, and the EMA class-statistic sharing is not covered by the theory and may not achieve consensus under the stated hyperparameters. These issues need to be resolved before the empirical and theoretical claims can be fully accepted.

major comments (4)
  1. [Section IV-B, Eqs. (9)-(10)] The adversarial training objective is specified inconsistently. L_adv in Eq. (9) minimizes the negative log-likelihood of the true class, which should make zl class-discriminative, while L^u_adv in Eq. (10) pushes the classifier toward uniform outputs over all classes. The paper does not define a min-max game, a gradient reversal layer, or alternating updates for the classifier parameters, so the two terms directly conflict as written. The claim that Learngene learns class-invariant representations is therefore not supported by the stated objective. Please clarify the actual training procedure and report the exact loss used in the experiments.
  2. [Appendix D, Lemma 2, Eq. (27)] The key step in Lemma 2 is the approximation Delta L ≈ η ||φ(t+1)E+0 − φ(t+1)E||², asserted without proof in step (a) of Eq. (27). This is not a consequence of Assumptions 1-3, and the loss change after aggregation is not generally proportional to the squared parameter displacement. Since Lemma 2 underpins Theorem 1 and Theorem 2, the claimed O(1/T) non-convex convergence rate is not established. Please either provide a correct proof under explicit assumptions on the Learngene loss module, or weaken the claim and present the analysis as a heuristic.
  3. [Algorithm 1, lines 6-10, and Theorem 2] The convergence analysis does not cover the EMA dynamics of class statistics. Algorithm 1 line 6 updates mu and Sigma with α=0.99, and lines 9-10 train PersonaNet using those statistics, but Assumption 3 only bounds the Learngene parameter variation; no analogue is provided for ||mu_tilde − mu|| or ||Sigma_tilde − Sigma||. Thus Theorem 2 does not apply to the full algorithm. Moreover, on a directed 20-client ring with α=0.99, the mixing rate of the EMA update is very slow: the second eigenvalue of 0.99I + 0.01S has magnitude about 0.9995, so after 300 rounds the non-uniform mode decays only to about 0.865 and the 'global' class statistics remain far from consensus. Please provide a mixing analysis, choose a hyperparameter that actually mixes, or supply direct experimental evidence that the shared statistics converge.
  4. [Tables I-II and Figures 3-6] No standard deviations, seeds, or significance tests are reported. Several of the claimed improvements over baselines are small (e.g., 0.11-0.22 points in some Global-T columns of Table I), and the abstract's 'up to 3.28%' claim rests on a single run. Without multiple seeds and error bars, the central empirical claim that DRDFL 'outperforms state-of-the-art methods' is not statistically verified. Please rerun with at least three random seeds and report mean±std, and perform a paired test for the main comparisons.
minor comments (5)
  1. [Section VII-B] The sentence 'Following [57]-[59], we set the parameter α in EMA to 0.99' cites time-series references [57] and [58] and the authors' own [59], none of which clearly justify the EMA momentum choice for federated class statistics. Please cite a more relevant source or provide a brief justification.
  2. [Figure 6] The label 'DDRFL' in Figure 6 should be 'DRDFL'.
  3. [Algorithm 1, line 2] The word 'pamrameter' is a typo for 'parameter'.
  4. [Section VII-C] The text 'CIFAR-100 setting with α = 0.1 and s = 30' should read 'β = 0.1' instead of 'α = 0.1'.
  5. [Section I and abstract] The distinction between the Learngene paradigm and the Learngene module is confusing after the footnote; consider using a different name for the module to avoid ambiguity throughout the paper.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the reported gains are external-benchmark accuracy numbers; the self-citations supply vocabulary and hyperparameters, not the predicted results.

full rationale

The paper's central empirical claims (Local-T/Global-T accuracy on SVHN, CIFAR-10, and CIFAR-100 against eight baselines, and 0.58 MB per-round communication) are measured on held-out test sets with standard baselines; no fitted constant is renamed as a prediction. The losses in Eqs. 6-12 define training objectives and do not contain the reported accuracies as inputs. The convergence proof (Theorems 1-2) is conditional on Lipschitz smoothness, bounded variance, and bounded Learngene variation (Assumptions 1-3); it does not assume the conclusion. The Learngene concept is cited from the authors' own prior work ([23]-[25], [41]-[48]), and the value α=0.99 in the EMA update is justified in part by the authors' prior paper [59], but these are motivational or hyperparameter choices, not load-bearing reductions of the measured outcomes. The reader's concern that EMA mixing on a 20-client ring is slow is a correctness or falsifiability issue about whether the method achieves the claimed global sharing; it is not an instance of a prediction being equivalent to its input by construction. Overall, the derivation chain is self-contained with respect to the empirical benchmarks.

Assumptions & free parameters 2 free parameters · 7 assumptions · 0 invented entities

The method rests on several unverified modeling assumptions: Gaussian class-conditional latents, a uniform-prior adversarial objective, and EMA-based ring propagation of global statistics. The convergence proof adds three standard assumptions plus an ad hoc bounded-variation assumption for Learngene. No new physical entities are introduced.

free parameters (2)
  • EMA momentum alpha = 0.99
    Set by hand in momentum updates of global class means and covariances (Sec. IV-A and Appendix VII-B); controls how much global statistics change when a neighbor's statistics arrive.
  • Reconstruction noise variance sigma^2 = not specified
    Gaussian noise n ~ N(0, sigma^2 I) is added to reconstructed samples in Sec. IV-C, and the paper says it improves robustness, but the value is not reported.
assumptions (7)
  • standard math Client loss functions are L1-Lipschitz smooth (Assumption 1).
    Used to derive the descent lemmas in Appendix D; this is a standard smoothness condition for non-convex SGD analysis.
  • standard math Stochastic gradients are unbiased and have bounded variance (Assumption 2).
    Standard SGD assumption used throughout the convergence proof.
  • ad hoc to paper Learngene parameter variation before and after ring averaging is bounded by delta^2 (Assumption 3).
    Introduced specifically to make Lemma 2 work; no experiment verifies the bound.
  • domain assumption A uniform class prior p_u(y=k)=1/K is the right target for global invariant knowledge.
    Adopted in Sec. III-C to handle label distribution skew; reasonable but not proven to be optimal.
  • domain assumption PersonaNet latent codes follow a Gaussian mixture distribution p(zp)=sum_k N(zp; mu_k, Sigma_k) p(k).
    Core modeling assumption in Sec. IV-A; class features are assumed Gaussian, and the learned KL objective depends on it.
  • domain assumption Minimizing the adversarial losses Ladv and Lu_adv with respect to the Learngene produces class-invariant representations.
    Stated in Sec. IV-B, but the losses as written do not form a standard minimax game and may have the opposite effect; this is the paper's weakest mechanistic premise.
  • domain assumption EMA updates with alpha=0.99 propagate stable global class statistics around the ring.
    Algorithm 1 lines 6-7; if the ring propagation does not converge, the PersonaNet KL target drifts.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Harmonizing Generalization and Personalization in Ring-topology Decentralized Federated Learning." pith.science (2026). https://pith.science/paper/CSR4QKUY

@misc{pith2026250419103,
  author       = {Pith},
  title        = {Pith review of: Harmonizing Generalization and Personalization in Ring-topology Decentralized Federated Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CSR4QKUY}},
  note         = {Machine review of arXiv:2504.19103}
}
read the original abstract

We introduce Ring-topology Decentralized Federated Learning (RDFL) for distributed model training, aiming to avoid the inherent risks of centralized failure in server-based FL. However, RDFL faces the challenge of low information-sharing efficiency due to the point-to-point communication manner when handling inherent data heterogeneity. Existing studies to mitigate data heterogeneity focus on personalized optimization of models, ignoring that the lack of shared information constraints can lead to large differences among models, weakening the benefits of collaborative learning. To tackle these challenges, we propose a Divide-and-conquer RDFL framework (DRDFL) that uses a feature generation model to extract personalized information and invariant shared knowledge from the underlying data distribution, ensuring both effective personalization and strong generalization. Specifically, we design a \textit{PersonaNet} module that encourages class-specific feature representations to follow a Gaussian mixture distribution, facilitating the learning of discriminative latent representations tailored to local data distributions. Meanwhile, the \textit{Learngene} module is introduced to encapsulate shared knowledge through an adversarial classifier to align latent representations and extract globally invariant information. Extensive experiments demonstrate that DRDFL outperforms state-of-the-art methods in various data heterogeneity settings.

Figures

Figures reproduced from arXiv: 2504.19103 by the authors.

Figure 1
Figure 1. Illustration of the optimized learning mechanism in RDFL and comparison of the personalization and generalization [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Illustration of the DRDFL framework. It adopts a [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Comparison of Local-T curves for different methods under various Non-IID settings on the CIFAR-10 dataset. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: The Non-IID data distribution simulated on different clients based on the CIFAR-10 dataset within the RDfl architecture. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Comparison of personalized performance curves for different methods under various Non-IID settings on the CIFAR-100 [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Comparison of Local-T and Global-T curves for different personalized methods on CIFAR-10 with [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Visualization of bird class samples from CIFAR-10 for different clients. The PersonaNet column presents the Grad￾CAM [64] outputs generated by the PersonaNet module, while the Learngene column illustrates the shared general representations captured by the Learngene mod…
Figure 8
Figure 8. Figure 8: Visualization the average performance of new clients joining the ring-topology federated learning system on CIFAR-10. [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: Visualization of the image reconstruction process based on leaked information, along with the difference values (Peak [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

66 extracted references · 46 canonical work pages

  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,” in Artificial Intelligence and Statistics , 2017, pp. 1273–1282

    2017

  2. [2]

    Trading off privacy, utility, and efficiency in federated learning,

    X. Zhang, Y . Kang, K. Chen, L. Fan, and Q. Yang, “Trading off privacy, utility, and efficiency in federated learning,” ACM Transactions on Intelligent Systems and Technology, vol. 14, no. 6, pp. 1–32, 2023

    work page 2023

  3. [3]

    Cross-silo prototypical calibration for federated learning with non-iid data,

    Z. Qi, L. Meng, Z. Chen, H. Hu, H. Lin, and X. Meng, “Cross-silo prototypical calibration for federated learning with non-iid data,” in Proceedings of the 31st ACM International Conference on Multimedia , 2023, pp. 3099–3107

    2023

  4. [4]

    Rethinking architecture design for tackling data heterogeneity in federated learning,

    L. Qu, Y . Zhou, P. P. Liang, Y . Xia, F. Wang, E. Adeli, L. Fei-Fei, and D. Rubin, “Rethinking architecture design for tackling data heterogeneity in federated learning,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2022, pp. 10 061–10 071

    work page 2022

  5. [5]

    Fedfed: Feature distillation against data heterogeneity in federated learning,

    Z. Yang, Y . Zhang, Y . Zheng, X. Tian, H. Peng, T. Liu, and B. Han, “Fedfed: Feature distillation against data heterogeneity in federated learning,” Advances in Neural Information Processing Systems , vol. 36, 2024

    2024

  6. [6]

    Towards efficient replay in federated incremental learning,

    Y . Li, Q. Li, H. Wang, R. Li, W. Zhong, and G. Zhang, “Towards efficient replay in federated incremental learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 12 820–12 829

    work page 2024

  7. [7]

    Cross-silo feature space alignment for federated learning on clients with imbalanced data,

    Z. Qi, L. Meng, Z. Li, H. Hu, and X. Meng, “Cross-silo feature space alignment for federated learning on clients with imbalanced data,” 2025

    work page 2025

  8. [8]

    Personalized federated domain-incremental learning based on adaptive knowledge matching,

    Y . Li, W. Xu, H. Wang, Y . Qi, J. Guo, and R. Li, “Personalized federated domain-incremental learning based on adaptive knowledge matching,” in European Conference on Computer Vision . Springer, 2025, pp. 127–144

    work page 2025

Show all 66 references
  1. [9]

    Ditto fair and robust federated learning through personalization,

    T. Li, S. Hu, A. Beirami, and V . Smith, “Ditto fair and robust federated learning through personalization,” in International Conference on Machine Learning. PMLR, 2021, pp. 6357–6368

    2021

  2. [10]

    How to prevent the poor performance clients for personalized federated learning?

    Z. Qu, X. Li, X. Han, R. Duan, C. Shen, and L. Chen, “How to prevent the poor performance clients for personalized federated learning?” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 12 167–12 176

    2023

  3. [11]

    On bridging generic and personalized federated learning for image classification,

    H.-Y . Chen and W.-L. Chao, “On bridging generic and personalized federated learning for image classification,” arXiv preprint arXiv2107.00778 , 2021

    2021 arXiv

  4. [12]

    Exploiting shared representations for personalized federated learning,

    L. Collins, H. Hassani, A. Mokhtari, and S. Shakkottai, “Exploiting shared representations for personalized federated learning,” in International Conference on Machine Learning , 2021, pp. 2089–2099

    2021

  5. [13]

    Fedbabu: Towards enhanced representation for federated image classification,

    J. Oh, S. Kim, and S.-Y . Yun, “Fedbabu: Towards enhanced representation for federated image classification,” arXiv preprint arXiv:2106.06042 , 2021

    2021 arXiv

  6. [14]

    Fedcp: Separating feature information for personalized federated learning via conditional policy,

    J. Zhang, Y . Hua, H. Wang, T. Song, Z. Xue, R. Ma, and H. Guan, “Fedcp: Separating feature information for personalized federated learning via conditional policy,” in Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2023, pp. 3249–3261

    2023

  7. [15]

    Perada: Parameter-efficient federated learning personalization with generalization guarantees,

    C. Xie, D.-A. Huang, W. Chu, D. Xu, C. Xiao, B. Li, and A. Anandkumar, “Perada: Parameter-efficient federated learning personalization with generalization guarantees,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 23 838–23 848

    2024

  8. [16]

    Dispfl: Towards communication-efficient personalized federated learning via decentralized sparse training,

    R. Dai, L. Shen, F. He, X. Tian, and D. Tao, “Dispfl: Towards communication-efficient personalized federated learning via decentralized sparse training,” in International conference on machine learning . PMLR, 2022, pp. 4587–4604

    2022

  9. [17]

    On the effectiveness of partial variance reduction in federated learning with heterogeneous data,

    B. Li, M. N. Schmidt, T. S. Alstrøm, and S. U. Stich, “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 , 2023, pp. 3964–3973

    2023

  10. [18]

    Decentralized directed collaboration for personalized federated learning,

    Y . Liu, Y . Shi, Q. Li, B. Wu, X. Wang, and L. Shen, “Decentralized directed collaboration for personalized federated learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 23 168–23 178

    2024

  11. [19]

    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 et al., “Advances and open problems in federated learning,” Foundations and Trendstextregistered in Machine Learning , vol. 14, no. 1–2, pp. 1–210, 2021

    2021

  12. [20]

    Learning to collaborate in decentralized learning of personalized models,

    S. Li, T. Zhou, X. Tian, and D. Tao, “Learning to collaborate in decentralized learning of personalized models,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 9766–9775

    2022

  13. [21]

    Personalized decentralized federated learning with knowledge distillation,

    E. Jeong and M. Kountouris, “Personalized decentralized federated learning with knowledge distillation,” in ICC 2023-IEEE International Conference on Communications. IEEE, 2023, pp. 1982–1987

    2023

  14. [22]

    Unleashing the power of continual learning on non-centralized devices: A survey,

    Y . Li, H. Wang, W. Xu, T. Xiao, H. Liu, M. Tu, Y . Wang, X. Yang, R. Zhang, S. Yu et al., “Unleashing the power of continual learning on non-centralized devices: A survey,” arXiv preprint arXiv:2412.13840 , 2024

    2024 arXiv

  15. [23]

    Learngene from open-world to your learning task,

    Q.-F. Wang, X. Geng, S.-X. Lin, S.-Y . Xia, L. Qi, and N. Xu, “Learngene from open-world to your learning task,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 36, no. 8, 2022, pp. 8557–8565

    2022

  16. [24]

    Learngene inheriting condensed knowledge from the ancestry model to descendant models,

    Q. Wang, X. Yang, S. Lin, and X. Geng, “Learngene inheriting condensed knowledge from the ancestry model to descendant models,” arXiv preprint arXiv2305.02279, pp. 1–17, 2023

    2023 arXiv

  17. [25]

    Initializing variable-sized vision transformers from learngene with learnable transformation,

    S. Xia, Y . Zu, X. Yang, and X. Geng, “Initializing variable-sized vision transformers from learngene with learnable transformation,” in Neural Information Processing Systems, 2024, pp. 1–26

    2024

  18. [26]

    Learn from others and be yourself in heterogeneous federated learning,

    W. Huang, M. Ye, and B. Du, “Learn from others and be yourself in heterogeneous federated learning,” in Proceedings of the IEEECVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 10 143–10 153

    2022

  19. [27]

    Fedproto federated prototype learning across heterogeneous clients,

    Y . Tan, G. Long, L. Liu, T. Zhou, Q. Lu, J. Jiang, and C. Zhang, “Fedproto federated prototype learning across heterogeneous clients,” in AAAI Conference on Artificial Intelligence , vol. 1, 2022, pp. 3–19

    2022

  20. [28]

    Fedtgp: Trainable global prototypes with adaptive-margin-enhanced contrastive learning for data and model heterogeneity in federated learning,

    J. Zhang, Y . Liu, Y . Hua, and J. Cao, “Fedtgp: Trainable global prototypes with adaptive-margin-enhanced contrastive learning for data and model heterogeneity in federated learning,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 38, no. 15, 2024, pp...

    2024

  21. [29]

    Recovering labels from local updates in federated learning,

    H. Chen and H. Vikalo, “Recovering labels from local updates in federated learning,” arXiv preprint arXiv:2405.00955 , 2024

    2024 arXiv

  22. [30]

    Decentralized federated learning: Balancing communication and computing costs,

    W. Liu, L. Chen, and W. Zhang, “Decentralized federated learning: Balancing communication and computing costs,” IEEE Transactions on Signal and Information Processing over Networks , vol. 8, pp. 131–143, 2022

    2022

  23. [31]

    Decentralized federated averaging,

    T. Sun, D. Li, and B. Wang, “Decentralized federated averaging,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 45, no. 4, pp. 4289–4301, 2022

    2022

  24. [32]

    Disentangled representation learning,

    X. Wang, H. Chen, Z. Wu, W. Zhu et al., “Disentangled representation learning,” IEEE Transactions on Pattern Analysis and Machine Intelligence , pp. 1–29, 2024

    2024

  25. [33]

    Commutative lie group vae for disentanglement learning,

    X. Zhu, C. Xu, and D. Tao, “Commutative lie group vae for disentanglement learning,” in International Conference on Machine Learning . PMLR, 2021, pp. 12 924–12 934

    2021

  26. [34]

    Dualvae: Dual disentangled variational autoencoder for recommendation,

    Z. Guo, G. Li, J. Li, C. Wang, and S. Shi, “Dualvae: Dual disentangled variational autoencoder for recommendation,” in Proceedings of the 2024 SIAM International Conference on Data Mining (SDM) . SIAM, 2024, pp. 571–579

    2024

  27. [35]

    Auto-encoding variational bayes,

    D. P. Kingma, “Auto-encoding variational bayes,” arXiv preprint arXiv:1312.6114 , 2013

    2013 arXiv

  28. [36]

    Learning structured output representation using deep conditional generative models,

    K. Sohn, H. Lee, and X. Yan, “Learning structured output representation using deep conditional generative models,” Advances in neural information processing systems, vol. 28, 2015

    2015

  29. [37]

    Personalization disentanglement for federated learning,

    P. Yan and G. Long, “Personalization disentanglement for federated learning,” in 2023 IEEE International Conference on Multimedia and Expo (ICME) . IEEE Computer Society, 2023, pp. 318–323

    2023

  30. [38]

    Disentangled federated learning for tackling attributes skew via invariant aggregation and diversity transferring,

    Z. Luo, Y . Wang, Z. Wang, Z. Sun, and T. Tan, “Disentangled federated learning for tackling attributes skew via invariant aggregation and diversity transferring,” in International Conference on Machine Learning . PMLR, 2022, pp. 14 527–14 541

    2022

  31. [39]

    On disentanglement of asymmetrical knowledge transfer for modality-task agnostic federated learning,

    J. Chen and A. Zhang, “On disentanglement of asymmetrical knowledge transfer for modality-task agnostic federated learning,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 38, no. 10, 2024, pp. 11 311–11 319

    2024

  32. [40]

    Spatio-temporal heterogeneous federated learning for time series classification with multi-view orthogonal training,

    C. Wu, H. Wang, X. Zhang, Z. Fang, and J. Bu, “Spatio-temporal heterogeneous federated learning for time series classification with multi-view orthogonal training,” in Proceedings of the 32nd ACM International Conference on Multimedia , 2024, pp. 2613–2622. JOURNAL OF LATEX CL...

    2024

  33. [41]

    Linearly decomposing and recomposing vision transformers for diverse-scale models,

    S. Lin, M. Zhang, R. Chen, X. Yang, Q. Wang, and X. Geng, “Linearly decomposing and recomposing vision transformers for diverse-scale models,” in The Thirty-eighth Annual Conference on Neural Information Processing Systems , 2024, pp. 1–25

    2024

  34. [42]

    Transformer as linear expansion of learngene,

    S. Xia, M. Zhang, X. Yang, R. Chen, H. Chen, and X. Geng, “Transformer as linear expansion of learngene,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 38, no. 14, 2024, pp. 16 014–16 022

    2024

  35. [43]

    Vision transformers as probabilistic expansion from learngene,

    Q. Wang, X. Yang, H. Chen, and X. Geng, “Vision transformers as probabilistic expansion from learngene,” in Forty-first International Conference on Machine Learning, 2024, pp. 1–14

    2024

  36. [44]

    Facilitating ai-based csi feedback deployment in massive mimo systems with learngene,

    X. Li, J. Guo, C.-K. Wen, X. Geng, and S. Jin, “Facilitating ai-based csi feedback deployment in massive mimo systems with learngene,” IEEE Transactions on Wireless Communications , 2024

    2024

  37. [45]

    Wave: Weight template for adaptive initialization of variable-sized models,

    F. Feng, Y . Xie, J. Wang, and X. Geng, “Wave: Weight template for adaptive initialization of variable-sized models,” arXiv preprint arXiv:2406.17503 , 2024

    2024 arXiv

  38. [46]

    Kind: Knowledge integration and diversion in diffusion models,

    Y . Xie, F. Feng, J. Wang, X. Geng, and Y . Rui, “Kind: Knowledge integration and diversion in diffusion models,” arXiv preprint arXiv:2408.07337 , 2024

    2024 arXiv

  39. [47]

    Building variable-sized models via learngene pool,

    B. Shi, S. Xia, X. Yang, H. Chen, Z. Kou, and X. Geng, “Building variable-sized models via learngene pool,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 38, no. 13, 2024, pp. 14 946–14 954

    2024

  40. [48]

    Transferring core knowledge via learngenes,

    F. Feng, J. Wang, and X. Geng, “Transferring core knowledge via learngenes,” arXiv preprint arXiv2401.08139 , 2024

    2024 arXiv

  41. [49]

    Rethinking feature distribution for loss functions in image classification,

    W. Wan, Y . Zhong, T. Li, and J. Chen, “Rethinking feature distribution for loss functions in image classification,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2018, pp. 9117–9126

    2018

  42. [50]

    Disentangling latent space for vae by label relevant/irrelevant dimensions,

    Z. Zheng and L. Sun, “Disentangling latent space for vae by label relevant/irrelevant dimensions,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2019, pp. 12 192–12 201

    2019

  43. [51]

    Infogan: Interpretable representation learning by information maximizing generative adversarial nets,

    X. Chen, Y . Duan, R. Houthooft, J. Schulman, I. Sutskever, and P. Abbeel, “Infogan: Interpretable representation learning by information maximizing generative adversarial nets,” Advances in neural information processing systems , vol. 29, 2016

    2016

  44. [52]

    Reading digits in natural images with unsupervised feature learning,

    Y . Netzer, T. Wang, A. Coates, A. Bissacco, B. Wu, A. Y . Ng et al., “Reading digits in natural images with unsupervised feature learning,” in NIPS workshop on deep learning and unsupervised feature learning , vol. 2011, no. 5. Granada, Spain, 2011, pp. 1–9

    2011

  45. [53]

    Learning multiple layers of features from tiny images,

    A. Krizhevsky, G. Hinton et al., “Learning multiple layers of features from tiny images,” pp. 1–60, 2009

    2009

  46. [54]

    Communication efficient primal-dual algorithm for nonconvex nonsmooth distributed optimization,

    C. Chen, J. Zhang, L. Shen, P. Zhao, and Z. Luo, “Communication efficient primal-dual algorithm for nonconvex nonsmooth distributed optimization,” in International conference on artificial intelligence and statistics . PMLR, 2021, pp. 1594–1602

    2021

  47. [55]

    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,” Advances in neural information processing systems , vol. 33, pp. 7611–7623, 2020

    2020

  48. [56]

    Fedbn: Federated learning on non-iid features via local batch normalization,

    X. Li, M. Jiang, X. Zhang, M. Kamp, and Q. Dou, “Fedbn: Federated learning on non-iid features via local batch normalization,” arXiv preprint arXiv:2102.07623, 2021

    2021 arXiv

  49. [57]

    A generalized fractionally integrated autoregressive moving-average process,

    C.-F. Chung, “A generalized fractionally integrated autoregressive moving-average process,” Journal of Time Series Analysis , vol. 17, no. 2, pp. 111–140, 1996

    1996

  50. [58]

    Following a trend with an exponential moving average: Analytical results for a gaussian model,

    D. S. Grebenkov and J. Serror, “Following a trend with an exponential moving average: Analytical results for a gaussian model,” Physica A: Statistical Mechanics and its Applications , vol. 394, pp. 288–303, 2014

    2014

  51. [59]

    Addressing skewed heterogeneity via federated prototype rectification with personalization,

    S. Guo, H. Wang, S. Lin, Z. Kou, and X. Geng, “Addressing skewed heterogeneity via federated prototype rectification with personalization,” IEEE Transactions on Neural Networks and Learning Systems , 2024

    2024

  52. [60]

    A unified federated learning framework for wireless communications: Towards privacy, efficiency, and security,

    H. Wen, Y . Wu, C. Yang, H. Duan, and S. Yu, “A unified federated learning framework for wireless communications: Towards privacy, efficiency, and security,” in IEEE INFOCOM 2020-IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS) . IEEE, 2020, pp. 653–658

    2020

  53. [61]

    Detecting malicious model updates from federated learning on conditional variational autoencoder,

    Z. Gu and Y . Yang, “Detecting malicious model updates from federated learning on conditional variational autoencoder,” in 2021 IEEE international parallel and distributed processing symposium (IPDPS) . IEEE, 2021, pp. 671–680

    2021

  54. [62]

    Data-free one-shot federated learning under very high statistical heterogeneity,

    C. E. Heinbaugh, E. Luz-Ricca, and H. Shao, “Data-free one-shot federated learning under very high statistical heterogeneity,” in The Eleventh International Conference on Learning Representations , 2023

    2023

  55. [63]

    Communication efficient distributed learning using variational auto encoders,

    A. Kasturi, S. Vishwamithra, and C. Hota, “Communication efficient distributed learning using variational auto encoders,” in 2022 14th International Conference on COMmunication Systems & NETworkS (COMSNETS) . IEEE, 2022, pp. 744–748

    2022

  56. [64]

    Grad-cam: Visual explanations from deep networks via gradient-based localization,

    R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, and D. Batra, “Grad-cam: Visual explanations from deep networks via gradient-based localization,” in Proceedings of the IEEE international conference on computer vision , 2017, pp. 618–626

    2017

  57. [65]

    Deep leakage from gradients,

    L. Zhu, Z. Liu, and S. Han, “Deep leakage from gradients,” Advances in neural information processing systems , vol. 32, 2019

    2019

  58. [66]

    pfedmoe: Data-level personalization with mixture of experts for model-heterogeneous personalized federated learning,

    L. Yi, H. Yu, C. Ren, H. Zhang, G. Wang, X. Liu, and X. Li, “pfedmoe: Data-level personalization with mixture of experts for model-heterogeneous personalized federated learning,” arXiv preprint arXiv:2402.01350 , 2024

    2024 arXiv

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.