Pith. sign in

REVIEW 5 major objections 5 minor 40 references

The paper claims that in one-shot federated learning, sharing knowledge in representation space—by aligning channels with CKA and injecting only the most compatible features as residuals—outperforms parameter-space aggregation under domain

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-04 11:12 UTC pith:IDOZZBH6

load-bearing objection CRIP's channel-level CKA matching with residual injection is a genuinely new and promising idea for one-shot personalized FL, with large wins on DomainNet and PACS, but the 'consistently outperforms' claim is contradicted by Office-Home, and the theory doesn't cover the actual algorithm as implemented. the 5 major comments →

arxiv 2608.02222 v1 pith:IDOZZBH6 submitted 2026-08-03 cs.LG

CRIP: Channel Level Representation Injection for Personalized One-Shot Federated Learning

Zijian Jiang , Chaoli Sun , Handing Wang , Xilu Wang This is my paper
classification cs.LG
keywords one-shot federated learningdomain heterogeneityrepresentation-space personalizationchannel alignmentCKAresidual injectionfeature fusionfew-shot learning
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper argues that in one-shot federated learning, where clients communicate only once, the right place to share knowledge is not the weight space but the representation space. It proposes CRIP, which has each client forward a small local mini-batch through every other client's feature extractor, measures channel-wise similarity with CKA, and injects only the best-matched foreign channels as residual features. This gives each client a personalized model that keeps its own parameters untouched while borrowing semantically compatible features from others. On three domain-heterogeneous benchmarks, CRIP reports gains over local models and strong baselines, especially on heavily shifted domains. If correct, the result means training-free representation fusion can substitute for parameter aggregation in one-shot settings.

Core claim

CRIP claims that under severe domain heterogeneity, one-shot federated learning can be solved in representation space: each client uploads only its feature extractor, the server broadcasts all extractors, and at inference each client runs a small local mini-batch through the other extractors, computes channel-wise CKA scores to find the best-matching channel per source, and adds the averaged aligned activations as a residual to its own feature map. The paper reports that this training-free procedure beats local models and existing one-shot methods on DomainNet, PACS, and Office-Home, with the largest gains on the most shifted domains. The accompanying theoretical decomposition splits each re

What carries the argument

The central object is the channel-level CKA alignment score s_{k,i}(c,c') computed on a calibration batch from the target domain. CKA (centered kernel alignment) is a similarity measure between two sets of activation vectors. For each target channel, CRIP selects the single best-matched channel per source client, averages their activations, and injects the result as a residual: z + α·z̄. The mechanism works because fusion is restricted to the high-CKA region, where the paper assumes (Eq. 17) that higher scores imply smaller expected semantic discrepancy; the averaging then suppresses uncorrelated residuals while the untouched target parameters preserve the model's own discriminative structur

Load-bearing premise

That the CKA similarity between a target channel and a source channel, measured on a small local batch, reliably tracks how semantically close those channels are on the test distribution—so that injecting only the top-matched channels adds transferable knowledge rather than noise; this is Eq. (17), an assumption the paper validates empirically on the three benchmarks rather than proves.

What would settle it

On a held-out domain pair, rank all channel pairs by CKA and compute their class-profile correlation Sim(c,c') as in Section V-C. If the top-1% CKA pairs have average Sim at or below zero, Eq. (17) fails for that pair, and CRIP should reduce accuracy below the target client's local model. Comparing CRIP with and without the alignment module on such a pair would directly test whether the claimed semantic-alignment mechanism carries the gain.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • One-shot federated learning can improve over local models even with no public dataset and no parameter averaging, as long as clients share feature extractors.
  • Under domain heterogeneity, representation-space alignment preserves transferable semantics better than parameter-space methods.
  • CKA-based channel selection is what makes fusion work: without alignment, averaging all channels degrades performance across all tested domains.
  • Fusing more source domains yields larger gains, consistent with a denoising effect from averaging uncorrelated residuals.
  • The method remains effective as the number of clients per domain grows, showing no obvious negative transfer at the tested scales.

Where Pith is reading between the lines

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

  • The method's success depends on clients having a small calibration batch from their own domain; in practice this implies a small labeled or unlabeled local set, which many FL deployments already possess but the paper does not discuss as a requirement.
  • CKA-based channel matching could serve as a general tool for cross-model transfer beyond FL, such as model merging or zero-shot knowledge distillation, wherever feature extractors see the same inputs.
  • Because inference cost grows linearly with the number of source clients, selecting a subset of high-CKA clients or parallelizing forward passes is a natural practical extension the paper does not develop.
  • If the alignment assumption holds only for channels with high CKA, a domain with unusual statistics (e.g., medical images) might need per-domain validation of Eq. (17) before injection.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes CRIP, a one-shot federated learning method for domain-heterogeneous clients. Each client uploads only its feature extractor; the server broadcasts all extractors back. For a target client, CRIP computes channel-wise CKA between target and source feature maps on a small local calibration batch, selects the highest-CKA channel from each source model for every target channel, and fuses the aligned source features into the target feature map via residual injection, leaving all parameters unchanged. Experiments on DomainNet, PACS, and Office-Home under a few-shot setting report strong average accuracies, and a theoretical section decomposes features into semantic and residual components to argue that the fusion is denoising under a CKA-based sufficient condition.

Significance. If the empirical claims could be substantiated, CRIP would be a valuable contribution: it is a representation-space, public-data-free approach to one-shot FL, and the channel-level CKA alignment idea is simple, training-free, and potentially generalizable. The paper also includes a useful ablation (CRIP-noA), sensitivity analyses, a scalability study, and a membership-inference evaluation, which are strengths. However, the central claim of consistent state-of-the-art performance is contradicted by the paper's own Table I; the theoretical guarantee in Section IV does not cover the implemented selection rule; and the main hyperparameter is tuned on the same benchmarks used for evaluation. The contribution is promising but requires substantive revision before its claims can be accepted.

major comments (5)
  1. [§III-B, §IV-B, §V-C, §V-E] Eq. (17) assumes a sufficient condition only for s ≥ τ, and the text in §IV-B explicitly says 'Channels selected only from the high-alignment region, where s≥τ.' Yet the algorithm in §III-B selects, for each target channel, the single argmax-CKA channel from each source client, with no threshold τ. The CKA quantiles reported in §V-E (q0.5=0.315, q0.9=0.507) show that many pairwise scores fall in a range where Eq. (17) is not assumed to hold, so the theoretical denoising guarantee does not cover the actual selection rule. Table III validates top-1% vs bottom-1% CKA regions, not the argmax-selected set. The paper should either enforce a threshold τ, validate Eq. (17) on the actual selected channels, or substantially weaken the theoretical claims. This is load-bearing because low-scoring argmax matches could inject domain-specific noise, consistent with CRIP's underperformance on Office-Hom
  2. [§III-B Eq. (11) vs §IV-B Eq. (18)] The implementation uses ⁱˆz = z + α·̄z in Eq. (11), while the theory analyzes ̃Z = (1−α)Z + (α/m)ΣZ_i in Eq. (18). These are not equivalent transformations because Eq. (11) leaves the target coefficient at 1 and adds α times the source average, whereas Eq. (18) renormalizes target and source weights to sum to 1. Since the prediction head was trained on z, the global scaling factor (1+α) in Eq. (11) changes the operating point of the head. Equations (19)–(21) rely on the convex weights in Eq. (18), so the variance-reduction derivation does not apply to the implemented injection. The authors must either add a normalization step and report it, or re-derive the bounds for the actual formula used.
  3. [§IV-B Eq. (21)] Eq. (21) claims the cross term in the residual decomposition vanishes 'since E[R_i|Y] = 0.' Zero conditional mean does not imply zero conditional covariance; E[⟨R_k, R_i⟩|Y] = Cov(Z_k, Z_i|Y), which need not vanish without an additional independence or decorrelation assumption. Consequently, the claimed denoising effect is not established by the derivation as written. The following sentence — 'if residuals across clients are not perfectly correlated given Y, then averaging strictly reduces the residual energy' — is also incomplete: positive but imperfect correlation can still leave the averaged residual energy above the individual level. The authors should state the additional assumption or provide the full quadratic form with the covariance terms.
  4. [§V-E, Table I] The main hyperparameter α is selected on the evaluation benchmarks. Section V-E states that 'in all experiments, we set α=0.8' after examining sensitivity on DomainNet (Fig. 5) and PACS-Cartoon (Table II), and the calibration batch size is also studied on DomainNet (Fig. 6). No held-out validation split is described for these choices. This constitutes selection on the test data and inflates the reported accuracies. Because α controls the core contribution (α=0 corresponds to the local model), a fair evaluation must set α and B on a validation split or via cross-validation, and should report results for a range of α rather than a single value tuned on the same benchmarks.
  5. [Abstract, §I, Table I] The abstract and introduction claim that CRIP 'consistently outperforms' state-of-the-art baselines. The paper's own Table I contradicts this. On Office-Home, FOL achieves 76.23±0.51 versus CRIP's 70.77±1.14; on PACS Photo, FOL achieves 97.15±0.88 versus CRIP's 93.38±0.43; on DomainNet Real, FDRL achieves 96.99 versus CRIP's 96.41. The discussion acknowledges FOL's Office-Home advantage only as 'considerably less stable' on DomainNet, which does not address the direct comparison. The claim must be softened to 'competitive' or supported with aggregate statistical tests across benchmarks. As written, the central empirical claim is not established.
minor comments (5)
  1. [§V-E] The CRIP-K variant is introduced but not defined in Section III, and no table or figure reports its results. The sentence 'To justify the use of CRIP-K' refers to an unexplained variant; please clarify the method or remove the reference.
  2. [§V-B] Typo: 'Muti-round' should be 'Multi-round.' Also, Table II uses 'PACS-Cartoon' while Table I uses 'C' for the Cartoon domain; consistent naming would help.
  3. [§V-E] In the paragraph beginning 'Overall, these results show that CLIP is sensitive...', 'CLIP' appears to be a typo for 'CRIP.' This should be corrected.
  4. [§V-G] Typo: 'Tabel II' should be 'Table II' in the reference to the supplementary material.
  5. [§IV-B] The notation in Eq. (16) uses X_k for target-domain inputs, while the decomposition in Eqs. (12)–(14) writes S_k and R_k for a random representation Z_k. The paper would benefit from explicitly stating that S_i and S_k in Eq. (17) are evaluated on the same target-domain input distribution.

Circularity Check

0 steps flagged

No significant circularity: CRIP's claims are empirical, and its theoretical section is explicitly conditional on an assumption rather than reducing to its inputs.

full rationale

The paper's derivation chain is not circular. CRIP's core mechanism is defined independently: channel-level CKA scores (Eq. 9), argmax channel correspondence, and residual injection (Eqs. 10-11). The theoretical section explicitly states that Eq. (17) is a 'local sufficient condition' and an assumption, not a derived theorem, and it is supported by external correlation analyses in Section V-C rather than by assuming the benchmark outcome. The subsequent bound in Eqs. (18)-(20) follows algebraically from the L2 projection decomposition and Jensen's inequality; it does not presuppose CRIP's measured accuracy. The benchmark results are external evaluations on DomainNet, PACS, and Office-Home against independent baselines, not outputs of the derivation. The value α=0.8 is selected from sensitivity analyses on the same datasets, which is a benchmark-tuning concern, not a definitional equivalence or a fitted parameter renamed as a prediction. The skeptic's observation that the algorithm uses argmax channels without enforcing the threshold τ in Eq. (17) is a gap between the theoretical sufficient condition and the implemented procedure; it concerns soundness or correctness risk, not circularity. There are no load-bearing author self-citations: the paper does not rely on prior work by the same authors to justify its method. Therefore no circular step can be exhibited with the required quote-and-reduction evidence.

Axiom & Free-Parameter Ledger

2 free parameters · 6 axioms · 0 invented entities

The paper's central mechanism depends on an assumed relationship between CKA scores and semantic correspondence, a conditional-expectation decomposition that assumes domain-invariant task semantics, and hand-set hyperparameters. No new physical or conceptual entities are postulated.

free parameters (2)
  • α (residual injection strength) = 0.8
    Chosen after sensitivity analysis on DomainNet (Fig. 5) and PACS-Cartoon (Table II); controls Eq. (11) and Eq. (18). Not predicted by the theory.
  • Calibration batch size |B| = not explicitly stated in main text
    Channel alignment uses a local mini-batch; sensitivity is tested from 5 to 200 (Fig. 6), but the default used for main results is not stated.
axioms (6)
  • domain assumption Task semantics are domain-invariant: T⊥D_k | Y in Eq. (12).
    Implicit in the semantic decomposition; if the domain factor changes the label semantics, the semantic components S_k across clients are not comparable.
  • standard math Conditional-expectation orthogonality E[R_k|Y]=0 from Eq. (13)-(15).
    Follows from the definition R_k = Z_k - E[Z_k|Y]; standard Hilbert-space projection property.
  • domain assumption Source-model residuals are conditionally mean-zero/uncorrelated given Y so averaging denoises in Eq. (21).
    The cross terms in Eq. (21) are dropped because E[R_i|Y]=0 for each i; whether this holds when source features are computed on target-domain inputs is assumed, not proven.
  • ad hoc to paper Eq. (17): higher CKA scores imply smaller expected semantic discrepancy in the high-score region.
    This is the central assumption for selective fusion; it is validated empirically in Section V-C on the same benchmarks rather than derived.
  • domain assumption CKA between channel activations is a valid proxy for semantic channel correspondence.
    Used to select channel pairs; CKA is a known representation-similarity metric, but its semantic validity at channel level is supported only by the paper's own validation.
  • domain assumption All clients share the same architecture and alignment layer ℓ.
    Required for channel-wise alignment and residual addition in Eqs. (9)-(11); the layer index ℓ is never explicitly specified in the experiments.

pith-pipeline@v1.3.0-daily-deepseek · 15908 in / 13695 out tokens · 117978 ms · 2026-08-04T11:12:05.516008+00:00 · methodology

0 comments
read the original abstract

One-shot federated learning (OSFL) has emerged as a promising collaborative model learning framework with only a single round of communication, offering significant advantages in communication efficiency and privacy preservation. However, OSFL often faces inherent limitations under severe domain heterogeneity across clients due to the lack of iterative knowledge exchange. Most existing OSFL methods require an auxiliary public dataset for knowledge distillation or leverage statistical information for parameter-level aggregation, overlooking feature shift caused by domain heterogeneity. To address these challenges, we propose CRIP, a personalized OSFL framework that operates in the representation space via channel-level feature alignment. To achieve this, each client uploads its feature extractor to the server, which broadcasts all extractors back to every client. Since not all source clients share compatible feature distributions with the target client, indiscriminate fusion of cross-client features would introduce domain-specific noise. Therefore, CRIP effectively measures the channel-wise representational similarity between the target client and each source client on a small local mini-batch, and selectively fuses only the most compatible features. Extensive experiments on domain-heterogeneous benchmarks such as DomainNet, PACS, and Office-Home demonstrate that CRIP consistently outperforms local models and state-of-the-art baselines, validating the effectiveness of representation-space personalization under extreme domain heterogeneity.

Figures

Figures reproduced from arXiv: 2608.02222 by Chaoli Sun, Handing Wang, Xilu Wang, Zijian Jiang.

Figure 1
Figure 1. Figure 1: FL settings and representation vs. parameter space alignment. The [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of CRIP. The target local model (in green) and client models (in other colors) perform feature extraction to obtain feature maps. In the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Illustration of domain heterogeneity in DomainNet. Each row corre [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Ablation study on DomainNet comparing CRIP with and without the [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Sensitivity analysis of domain-specific samples [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 5
Figure 5. Figure 5: Sensitivity analysis of the residual injection strength [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Scalability analysis of the proposed method on DomainNet. The bar [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

40 extracted references · 1 canonical work pages · 1 internal anchor

  1. [1]

    A survey on federated learning: challenges and applications,

    J. Wen, Z. Zhang, Y . Lan, Z. Cui, J. Cai, and W. Zhang, “A survey on federated learning: challenges and applications,”International journal of machine learning and cybernetics, vol. 14, no. 2, pp. 513–535, 2023

    2023

  2. [2]

    Fedbens: One-shot federated learning based on bayesian ensemble,

    J. Talpini, M. Savi, and G. Neglia, “Fedbens: One-shot federated learning based on bayesian ensemble,” inForty-second International Conference on Machine Learning

  3. [3]

    Fedtmos: Efficient one-shot federated learning with tsetlin machine,

    S. H. S. Qi, J. Chauhan, G. V . Merrett, and J. Hare, “Fedtmos: Efficient one-shot federated learning with tsetlin machine,” inThe Thirteenth International Conference on Learning Representations, 2025

    2025

  4. [4]

    Revisiting ensembling in one-shot federated learning,

    Y . Allouah, A. Dhasade, R. Guerraoui, N. Gupta, A.-M. Kermarrec, R. Pinot, R. Pires, and R. Sharma, “Revisiting ensembling in one-shot federated learning,”Advances in Neural Information Processing Systems, vol. 37, pp. 68 500–68 527, 2024

    2024

  5. [5]

    One-shot feder- ated learning: theoretical limits and algorithms to achieve them,

    S. Salehkaleybar, A. Sharifnassab, and S. J. Golestani, “One-shot feder- ated learning: theoretical limits and algorithms to achieve them,”Journal of Machine Learning Research, vol. 22, no. 189, pp. 1–47, 2021

    2021

  6. [6]

    Dense: Data-free one-shot federated learning,

    J. Zhang, C. Chen, B. Li, L. Lyu, S. Wu, S. Ding, C. Shen, and C. Wu, “Dense: Data-free one-shot federated learning,”Advances in Neural Information Processing Systems, vol. 35, pp. 21 414–21 428, 2022

    2022

  7. [7]

    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,” inThe Eleventh International Conference on Learning Representations, 2023

    2023

  8. [8]

    Capture global feature statistics for one-shot federated learning,

    Z. Guan, Y . Zhou, and X. Gu, “Capture global feature statistics for one-shot federated learning,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 39, no. 16, 2025, pp. 16 942–16 950

    2025

  9. [9]

    Federated learning via decentralized dataset distillation in resource-constrained edge environments,

    R. Song, D. Liu, D. Z. Chen, A. Festag, C. Trinitis, M. Schulz, and A. Knoll, “Federated learning via decentralized dataset distillation in resource-constrained edge environments,” in2023 International Joint Conference on Neural Networks (IJCNN), 2023, pp. 1–10

    2023

  10. [10]

    Enhancing one- shot federated learning through data and ensemble co-boosting,

    R. Dai, Y . Zhang, A. Li, T. Liu, X. Yang, and B. Han, “Enhancing one- shot federated learning through data and ensemble co-boosting,” inThe Twelfth International Conference on Learning Representations, 2024. [Online]. Available: https://openreview.net/forum?id=tm8s3696Ox

    2024

  11. [11]

    Federated oriented learning: A practical one-shot personalized federated learning framework,

    G. Huang and T. Shu, “Federated oriented learning: A practical one-shot personalized federated learning framework,” inForty-second International Conference on Machine Learning, 2025. [Online]. Available: https://openreview.net/forum?id=jwjvkWsePB

    2025

  12. [12]

    Fedbip: Heterogeneous one-shot federated learning with personalized latent diffusion models,

    H. Chen, H. Li, Y . Zhang, G. Zhang, J. Bi, P. Torr, J. Gu, D. Krompass, and V . Tresp, “Fedbip: Heterogeneous one-shot federated learning with personalized latent diffusion models,”CoRR, vol. abs/2410.04810,

    Pith/arXiv arXiv

  13. [13]

    Feddeo: Description-enhanced one-shot federated learning with diffusion models,

    M. Yang, S. Su, B. Li, and X. Xue, “Feddeo: Description-enhanced one-shot federated learning with diffusion models,” inProceedings of the 32nd ACM International Conference on Multimedia, 2024, pp. 6666– 6675

    2024

  14. [14]

    Federated generative learning with foundation models

    J. Zhang, X. hua Qi, S. Pang, S. Pan, X. Tu, P. Wan, and B. Zhao, “Federated generative learning with foundation models.”

  15. [15]

    Fedlpa: One- shot federated learning with layer-wise posterior aggregation,

    X. Liu, L. Liu, F. Ye, Y . Shen, X. Li, L. Jiang, and J. Li, “Fedlpa: One- shot federated learning with layer-wise posterior aggregation,”Advances in Neural Information Processing Systems, vol. 37, pp. 81 510–81 548, 2024

    2024

  16. [16]

    Fedfisher: Leveraging fisher information for one-shot federated learning,

    D. Jhunjhunwala, S. Wang, and G. Joshi, “Fedfisher: Leveraging fisher information for one-shot federated learning,” inInternational Conference on Artificial Intelligence and Statistics. PMLR, 2024, pp. 1612–1620

    2024

  17. [17]

    Fusefl: One-shot federated learning through the lens of causality with progressive model fusion,

    Z. Tang, Y . Zhang, P. Dong, Y .-m. Cheung, A. Zhou, B. Han, and X. Chu, “Fusefl: One-shot federated learning through the lens of causality with progressive model fusion,”Advances in Neural Information Processing Systems, vol. 37, pp. 28 393–28 429, 2024

    2024

  18. [18]

    Does one-shot give the best shot? mitigating model inconsistency in one-shot federated learning,

    H. Zeng, W. Huang, T. Zhou, X. Wu, G. Wan, Y . Chen, and Z. Cai, “Does one-shot give the best shot? mitigating model inconsistency in one-shot federated learning,” inForty-second International Conference on Machine Learning

  19. [19]

    A frequency-based approach for federated domain generalization in heterogeneous medical imaging,

    D. Wang, Y . Cui, Z. Wu, H. Xi, and J. Zhu, “A frequency-based approach for federated domain generalization in heterogeneous medical imaging,” inProceedings of the 2025 International Conference on Multimedia Retrieval, ser. ICMR ’25. New York, NY , USA: Association for Computing Machinery, 2025, p. 1313–1321. [Online]. Available: https://doi.org/10.1145/3...

    arXiv 2025

  20. [20]

    Fed2: Feature-aligned federated learning,

    F. Yu, W. Zhang, Z. Qin, Z. Xu, D. Wang, C. Liu, Z. Tian, and X. Chen, “Fed2: Feature-aligned federated learning,” inProceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining, 2021, pp. 2066–2074

    2021

  21. [21]

    Moment matching for multi-source domain adaptation,

    X. Peng, Q. Bai, X. Xia, Z. Huang, K. Saenko, and B. Wang, “Moment matching for multi-source domain adaptation,” inProceedings of the IEEE International Conference on Computer Vision, 2019, pp. 1406– 1415

    2019

  22. [22]

    pfedafm: Adaptive feature mixture for batch-level personalization in heterogeneous federated learning,

    L. Yi, H. Yu, C. Ren, H. Zhang, G. Wang, X. Liu, and X. Li, “pfedafm: Adaptive feature mixture for batch-level personalization in heterogeneous federated learning,”CoRR, 2024

    2024

  23. [23]

    Fedrda: Representation deviation alignment in heterogeneous federated learning,

    W. Yao, G. Sun, S. Zhu, R. Wang, X. Zhu, H. Xu, and X. Wei, “Fedrda: Representation deviation alignment in heterogeneous federated learning,”IEEE Transactions on Industrial Informatics, 2025

    2025

  24. [24]

    One-shot federated learning via synthetic distiller-distillate communication,

    J. Zhang, S. Liu, and X. Wang, “One-shot federated learning via synthetic distiller-distillate communication,”Advances in Neural Infor- mation Processing Systems, vol. 37, pp. 102 611–102 633, 2024

    2024

  25. [25]

    Distilled one-shot federated learning,

    Y . Zhou, G. Pu, X. Ma, X. Li, and D. Wu, “Distilled one-shot federated learning,”arXiv preprint arXiv:2009.07999, 2020

    Pith/arXiv arXiv 2009

  26. [26]

    Xor mixup: Privacy-preserving data augmentation for one-shot federated learning,

    M. Shin, C. Hwang, J. Kim, J. Park, M. Bennis, and S.-L. Kim, “Xor mixup: Privacy-preserving data augmentation for one-shot federated learning,”arXiv preprint arXiv:2006.05148, 2020

    Pith/arXiv arXiv 2006

  27. [27]

    Parametric feature transfer: One-shot federated learning with foundation models,

    M. Beitollahi, A. Bie, S. Hemati, L. M. Brunswic, X. Li, X. Chen, and G. Zhang, “Parametric feature transfer: One-shot federated learning with foundation models,”arXiv preprint arXiv:2402.01862, 2024

    Pith/arXiv arXiv 2024

  28. [28]

    Osgan: One-shot distributed learning using generative adversarial networks: A. kasturi, c. hota,

    A. Kasturi and C. Hota, “Osgan: One-shot distributed learning using generative adversarial networks: A. kasturi, c. hota,”The Journal of Supercomputing, vol. 79, no. 12, pp. 13 620–13 640, 2023

    2023

  29. [29]

    Clip-guided federated learning on heterogeneity and long-tailed data,

    J. Shi, S. Zheng, X. Yin, Y . Lu, Y . Xie, and Y . Qu, “Clip-guided federated learning on heterogeneity and long-tailed data,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 38, no. 13, 2024, pp. 14 955–14 963

    2024

  30. [30]

    Gpfl: Simultaneously learning global and personalized feature information for personalized federated learning,

    J. Zhang, Y . Hua, H. Wang, T. Song, Z. Xue, R. Ma, J. Cao, and H. Guan, “Gpfl: Simultaneously learning global and personalized feature information for personalized federated learning,” inProceedings of the IEEE/CVF international conference on computer vision, 2023, pp. 5041– 5051

    2023

  31. [31]

    Fraug: Tackling federated learning with non-iid features via representation augmenta- tion,

    H. Chen, A. Frikha, D. Krompass, J. Gu, and V . Tresp, “Fraug: Tackling federated learning with non-iid features via representation augmenta- tion,” inProceedings of the IEEE/CVF international conference on computer vision, 2023, pp. 4849–4859

    2023

  32. [32]

    Fedfa: Federated learning with feature anchors to align features and classifiers for heterogeneous data,

    T. Zhou, J. Zhang, and D. H. Tsang, “Fedfa: Federated learning with feature anchors to align features and classifiers for heterogeneous data,” IEEE Transactions on Mobile Computing, vol. 23, no. 6, pp. 6731–6742, 2023

    2023

  33. [33]

    Fedfm: Anchor- based feature matching for data heterogeneity in federated learning,

    R. Ye, Z. Ni, C. Xu, J. Wang, S. Chen, and Y . C. Eldar, “Fedfm: Anchor- based feature matching for data heterogeneity in federated learning,” IEEE Transactions on Signal Processing, vol. 71, pp. 4224–4239, 2023

    2023

  34. [34]

    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, pp. 60 397–60 428, 2023

    2023

  35. [35]

    Similarity of neural network representations revisited,

    S. Kornblith, M. Norouzi, H. Lee, and G. Hinton, “Similarity of neural network representations revisited,” inInternational conference on machine learning. PMlR, 2019, pp. 3519–3529

    2019

  36. [36]

    Deep domain- adversarial image generation for domain generalisation,

    K. Zhou, Y . Yang, T. Hospedales, and T. Xiang, “Deep domain- adversarial image generation for domain generalisation,” inProceedings of the AAAI conference on artificial intelligence, vol. 34, no. 07, 2020, pp. 13 025–13 032

    2020

  37. [37]

    Deep hashing network for unsupervised domain adaptation,

    H. Venkateswara, J. Eusebio, S. Chakraborty, and S. Panchanathan, “Deep hashing network for unsupervised domain adaptation,” inPro- ceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 5018–5027

    2017

  38. [38]

    Fed- erated discriminative representation learning for image classification,

    Y . Zhang, Y . Wang, Y . Li, Y . Xu, S. Wei, S. Liu, and X. Shang, “Fed- erated discriminative representation learning for image classification,” IEEE Transactions on Neural Networks and Learning Systems, vol. 36, no. 2, pp. 3204–3217, 2025

    2025

  39. [39]

    Vector quantization-based clustered federated learning with global feature anchors for improved representation and generalization,

    X. Chen, Y . Zhang, X. Xu, D. Hu, and G. Xu, “Vector quantization-based clustered federated learning with global feature anchors for improved representation and generalization,”IEEE Transactions on Neural Net- works and Learning Systems, vol. 36, no. 11, pp. 19 790–19 804, 2025

    2025

  40. [2024]

    FedBiP: Heterogeneous One-Shot Federated Learning with Personalized Latent Diffusion Models

    [Online]. Available: https://doi.org/10.48550/arXiv.2410.04810

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2410.04810