Pith. sign in

REVIEW 4 major objections 4 minor 33 references

Fed-REACT claims that separating a shared representation-learning phase from evolutionarily clustered task models handles both heterogeneous and drifting client data, reporting higher accuracy and lower error than supervised federated basel

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 ↗

Fed-REACT first trains a shared encoder, then repeatedly clusters clients by smoothed task-model weights, improving federated learning accuracy on heterogeneous, non-stationary time series.

T0 review reviewed 2026-08-04 challenge →

load-bearing objection Genuinely new combination of self-supervised learning and evolutionary clustering for federated time-series, but the SUMO headline is a test-set-selected optimum at full personalization, so the superiority claim needs an honest rework. the 4 major comments →

arxiv 2509.07198 v1 pith:QK2YPCMJ submitted 2025-09-08 cs.LG cs.DC

Fed-REACT: Federated Representation Learning for Heterogeneous and Evolving Data

Yiyue Chen , Usman Akram , Chianing Wang , Haris Vikalo This is my paper
classification cs.LG cs.DC
keywords federated learningself-supervised representation learningnon-IID dataconcept driftevolutionary clusteringtime-series classificationclient clusteringregret bound
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 reading

The paper tries to show that federated learning can cope with two difficulties at once: client data distributions differ from each other, and each client's distribution drifts over time. Its proposed method, Fed-REACT, works in two stages: first, clients jointly learn a shared time-series encoder from unlabeled data using contrastive self-supervised learning; second, the server clusters clients by the similarity of their task-model weights, using evolutionary clustering that smooths cluster assignments across rounds so drift does not cause spurious regroupings. The paper also gives a regret bound for the representation-learning stage and reports better accuracy and lower error than supervised federated baselines on the RTD, EEG, and SUMO datasets. This matters because real deployments in healthcare, driving, and finance routinely face both non-IID and time-varying data.

Core claim

The paper's central claim is that two forms of heterogeneity should be handled separately: low-level temporal features are sufficiently shared across clients to be learned by one global encoder, whereas task-specific behavior is tied to local distributions and should be learned by per-cluster models whose membership evolves over time. Fed-REACT operationalizes this with a contrastive encoder in phase 1 and, in phase 2, an adaptive evolutionary clustering scheme that estimates a smoothed client-similarity matrix from task-model weights and groups clients with agglomerative hierarchical clustering. The theoretical analysis shows that time-smoothed gradient descent on the encoder's simplified o

What carries the argument

The load-bearing mechanism is the adaptive evolutionary clustering step, an AFFECT-style update with an adaptive forgetting factor: the server maintains a smoothed similarity matrix ψ^t = a_t ψ^{t-1} + (1 - a_t)W_t, where W_t is the current cosine similarity between clients' vectorized task-model weights and a_t is estimated from the variance and drift of those similarities. This smoothing is what lets cluster memberships persist through transient noise and only change when the underlying distribution genuinely shifts. The representation phase is carried by a contrastive loss on time-series sub-sequences, and its theory uses a time-smoothed gradient descent update whose regret is bounded by

Load-bearing premise

The reported gains depend on an honest comparison: cluster count C must not be tuned on test labels (the paper reports doing so), and the unlabeled encoder data available to Fed-REACT must not be the true source of the advantage.

What would settle it

Run the SUMO and RTD experiments again with C chosen by a validation set (elbow or silhouette) rather than by best test performance, and give the supervised baselines the same self-supervised encoder pretraining; if Fed-REACT's accuracy or RMSE margin shrinks to baseline levels, its empirical claim is selection- or data-access-driven.

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

If this is right

  • With distribution drift, snapshot clustering misassigns clients; Fed-REACT's smoothed similarity matrix recovers ground-truth clusters within a few rounds and keeps them stable, as measured by Rand score.
  • The shared encoder is trained once on unlabeled data and reused for lightweight cluster-specific task models, reducing communication: the server sends each client one model per cluster rather than all cluster models.
  • The regret bound for time-smoothed gradient descent says that with step size 1/β and enough smoothing, the average gradient norm of the smoothed objective is controlled mainly by projection error, not by the drift itself.
  • The framework extends to partial client participation by reusing the most recent task-model weights for absent clients when computing the similarity matrix.
  • If the number of clusters is unknown, the paper's experiments show that an elbow method on within-cluster sum of squares can select the correct cluster count.

Where Pith is reading between the lines

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

  • An implicit consequence of the two-phase split is that part of the accuracy gain could come from the contrastive encoder seeing all unlabeled data, while supervised baselines see only labeled samples; ablating the encoder under identical labeled budgets would separate these two contributions.
  • Because Appendix H shows that SUMO's best RMSE occurs at C=50, full personalization, the paper's regression claim is effectively about representation quality rather than cluster coordination; similar patterns may appear in other highly heterogeneous tasks.
  • The adaptive forgetting factor estimates cluster stability from pairwise weight similarities, so in very large client populations the O(K^2) similarity matrix could become a bottleneck, suggesting a batched or sampled similarity estimator as a natural extension.
Share X Bluesky LinkedIn Reddit HN

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

4 major / 4 minor

Summary. The paper proposes Fed-REACT, a two-stage federated learning framework for heterogeneous and evolving time-series data. Phase 1 learns a shared representation encoder via self-supervised contrastive learning on local unlabeled data; Phase 2 clusters clients based on task-model weights using evolutionary clustering (AFFECT) with an adaptive forgetting factor, and aggregates cluster-specific task models using either simple temporal averaging or weighted averaging with forgetting. The authors provide a regret-based convergence analysis of a simplified linear SSL objective and report experiments on RTD, EEG, and SUMO, claiming superior accuracy and robustness over supervised and clustered FL baselines.

Significance. If the claims hold, the framework addresses a real deployment gap: federated learning under both inter-client heterogeneity and intra-client distribution drift. The combination of self-supervised representation learning with evolutionary clustering is a plausible and timely idea, and the paper provides concrete algorithms, an extensive appendix, and a theoretical framing. However, the empirical evidence for the headline claim is weakened by a test-set-based selection of the cluster count and by an unaddressed data-access asymmetry with the baselines, while the theoretical proof contains an invalid limit step. The significance is therefore conditional on these issues being resolved.

major comments (4)
  1. [Section 4.1 / Appendix H (Table 6)] The headline SUMO result (RMSE 1.3 vs 20.1 for the best baseline) is obtained by selecting the number of clusters C on the test set. Section 4.1 states that 'we search over a range of values to select the one yielding the best performance,' and Appendix H shows that the reported optimum is C=50, which the appendix itself describes as 'complete personalization.' Thus Table 1 reports the maximum of a test-set sweep over C, not an unbiased estimate; moreover, with C=50 the evolutionary clustering component of Fed-REACT is not exercised at all. Since this is the largest margin claimed in the paper, the central empirical claim is not supportable as presented. Please select C on a validation set (or via the Appendix E estimators) and report the corresponding test performance, or present the full sweep as the main result with per-C uncertainty.
  2. [Section 4.1 / Table 1] The comparison is confounded by data access. Fed-REACT's Phase 1 trains the encoder on all unlabeled local data, whereas the supervised baselines are trained only on labeled samples. With 90/10 train/test splits, the baselines presumably see a subset of the data that Fed-REACT also uses, plus labels; part of the reported improvement may therefore reflect additional unlabeled data rather than the method itself. Please include a semi-supervised or self-supervised baseline trained on the same unlabeled data, or restrict Phase 1 to the labeled training subset, so that the effect of representation learning is isolated.
  3. [Appendix C (Proof of Theorem 1)] The proof does not establish the stated limit. After substitution, the bound contains the term (32βM/W)·(2−γ^w+γ^{w−1})/(1−γ). As γ→1−, W→w and the numerator tends to 2, so this term diverges as 1/(1−γ); it does not converge to 64βM/W as claimed. The line 'lim_{γ→1−} ||∇S||² ≤ (1/W)(64βM+2σ²)+5/8ε²' therefore does not follow from the preceding algebra. A correct derivation must keep (1−γ^w)/(1−γ) terms together so that the finite limit emerges (the theorem may be true with different constants, but the proof as written is invalid). In addition, the theorem is stated as a bound on the random quantity ||∇S||² without an expectation, while the proof uses conditional expectation; the statement should be in expectation or with high probability.
  4. [Section 4 (Tables 1–3)] No error bars, standard deviations, or random seeds are reported for any experiment. The 'robustness' claim in the abstract and Section 4 cannot be assessed from single-point comparisons, especially when accuracies are reported to three decimals and differences between methods can be as small as 0.01. Please report means and standard errors over at least 3–5 seeds, and also report the variance of the Rand scores for the clustering experiments.
minor comments (4)
  1. [Section 2 / Algorithm 1] Notation is inconsistent: the problem setup uses n clients, while Algorithm 1 uses K for the number of clients and also overloads k as a client index. Please unify.
  2. [Section 4.2] The text says 'Figure 11 demonstrates...' but the referenced figure appears to be Figure 1 (and the caption numbering is otherwise out of order). Please correct the cross-reference.
  3. [Assumption 3.1(c)] The notation is confusing: ε_proj is defined as a vector, but the bound is written as ||ε_proj||₂ ≤ ε², which mixes a squared scalar with a norm. Please use, e.g., ||ε_proj|| ≤ ε or ||ε_proj||² ≤ ε² consistently.
  4. [Appendix H / Table 6] The statement that C=50 denotes 'complete personalization' is important context for interpreting the main result, but it is relegated to the appendix. This context should be stated in Section 4.1 when the SUMO result is presented, or the main table should include the selected C value.

Circularity Check

1 steps flagged

The theoretical derivation is self-contained, but the headline SUMO result is obtained by selecting the number of clusters on the test set (C=50, full personalization), making that particular 'prediction' an optimistic test-selected optimum rather than an unbiased evaluation.

specific steps
  1. fitted input called prediction [Section 4.1 (SUMO EV paragraph); Appendix H, Table 6]
    "Since the number of clusters C is not known in advance, we search over a range of values to select the one yielding the best performance. ... [Appendix H] the optimal performance is achieved for a completely personalized setting (C=50)."

    The reported SUMO RMSE of 1.3 is not the performance of a fixed Fed-REACT configuration; it is the minimum over a test-set sweep of C. Because C is selected by 'best performance' on the same test set, the headline number equals the best test value by construction (RMSE 1.3 = min_C test_RMSE(C)). This is a fitted hyperparameter renamed as a prediction: the evaluation target is used to choose the reported configuration, so the SUMO superiority claim is statistically forced rather than independently predicted. The paper is transparent in Appendix H, but Table 1 presents the selected value without noting it is a test-selected optimum.

full rationale

The core Fed-REACT derivation is not circular. Phase 1 is standard federated contrastive representation learning; Phase 2 adopts AFFECT evolutionary clustering (Xu et al. 2014) with two aggregation schemes. The theoretical regret bound (Theorem 1) follows from stated Lipschitz/smoothness/bounded-variance assumptions and the Aydore et al. time-smoothed gradient framework; the epsilon^2 floor is assumed in Assumption 3.1, not derived from the target result, so the bound is a normal worst-case analysis rather than a self-fulfilling identity. The self-citation to Fed-QSSL (Chen et al., 2024) appears only in related work and is not load-bearing. The main circularity concern is empirical: the strongest margin (SUMO, RMSE 1.3 vs ~20 for the best baseline) is obtained by choosing C on the test set, and Appendix H shows the optimum is C=50, i.e., complete personalization where evolutionary clustering is not exercised. Thus that specific 'prediction' reduces to a test-selected optimum by construction, making the empirical superiority claim partially circular/optimistic even though the algorithm design itself is not forced by self-citation. Correctness risks unrelated to circularity (e.g., Phase 1 using all unlabeled data while supervised baselines use only labeled data) are noted but not scored here.

Axiom & Free-Parameter Ledger

1 free parameters · 3 axioms · 0 invented entities

The main free parameter is the cluster count C, tuned on the test set for SUMO. The theoretical analysis relies on standard smoothness and noise assumptions plus an ad hoc simplification of the SSL objective to a linear PCA-like form. No new physical or conceptual entities are introduced.

free parameters (1)
  • Number of clusters C = 3 for RTD and EEG (known a priori); 50 for SUMO (selected by test-set performance)
    For SUMO, the paper searches over C and picks the value with lowest test RMSE (Appendix H, Table 6), making the reported 1.3 a test-selected result.
axioms (3)
  • domain assumption Assumption 3.1: loss bounded by M, L-Lipschitz and β-smooth, unbiased stochastic gradients with bounded variance σ², and projection error bounded by ε²
    Invoked in Section 3 to prove Theorem 1. Standard in optimization but not verified for the actual contrastive loss with a Causal CNN encoder.
  • ad hoc to paper The simplified linear SSL objective f(θ) = argmin_θ ||Xbar - θ^T θ||² is a faithful surrogate for the contrastive loss (1)
    Section 3 replaces the actual contrastive loss with a PCA-like objective and assumes a linear encoder, while the deployed algorithm uses a nonlinear Causal CNN and the original loss. The theory may not transfer to the actual algorithm.
  • domain assumption Task model weights reflect underlying client data distribution similarity
    Phase II clusters clients based on task model weights (Section 2). If the weights are noisy, high-variance, or non-informative, the cluster assignments and thus the cluster-wise models are not meaningful.

reviewed 2026-08-04 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Fed-REACT: Federated Representation Learning for Heterogeneous and Evolving Data." pith.science (2026). https://pith.science/paper/QK2YPCMJ

@misc{pith2026250907198,
  author       = {Pith},
  title        = {Pith review of: Fed-REACT: Federated Representation Learning for Heterogeneous and Evolving Data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QK2YPCMJ}},
  note         = {Machine review of arXiv:2509.07198}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Motivated by the high resource costs and privacy concerns associated with centralized machine learning, federated learning (FL) has emerged as an efficient alternative that enables clients to collaboratively train a global model while keeping their data local. However, in real-world deployments, client data distributions often evolve over time and differ significantly across clients, introducing heterogeneity that degrades the performance of standard FL algorithms. In this work, we introduce Fed-REACT, a federated learning framework designed for heterogeneous and evolving client data. Fed-REACT combines representation learning with evolutionary clustering in a two-stage process: (1) in the first stage, each client learns a local model to extracts feature representations from its data; (2) in the second stage, the server dynamically groups clients into clusters based on these representations and coordinates cluster-wise training of task-specific models for downstream objectives such as classification or regression. We provide a theoretical analysis of the representation learning stage, and empirically demonstrate that Fed-REACT achieves superior accuracy and robustness on real-world datasets.

Figures

Figures reproduced from arXiv: 2509.07198 by Chianing Wang, Haris Vikalo, Usman Akram, Yiyue Chen.

Figure 1
Figure 1. Figure 1: Rand scores for client clustering under distribution drift (RTD, EEG). We evaluate two settings: 10 clients grouped into 3 clusters over 100 rounds, and 100 clients across 3 clusters over 200 rounds. In both, each client receives |Mk t | = 64 training samples per round. Results are reported in [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Label distribution for three clusters generated using Dirichlet distribution with [PITH_FULL_IMAGE:figures/full_fig_p016_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: WCSS scores against the number of clusters [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Rand score against the ground truth for different method on the RTD dataset for intermittent client participa￾tion (Client Participation Ratio: 0.33) [PITH_FULL_IMAGE:figures/full_fig_p017_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Rand score against the ground truth for different method on the RTD dataset for stationary cluster distribu￾tions for a setup with 10 clients [PITH_FULL_IMAGE:figures/full_fig_p018_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: Rand score against the ground truth for different method on the RTD dataset with client migration setting [PITH_FULL_IMAGE:figures/full_fig_p018_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: The Rand score of Fed￾REACT vs. baseline methods on RTD for Non Stationary Setting (Strategy 1), 10 clients, |Mk t | = 64 training samples, Ttask = 100., λ1 = 0.75, λ2 = 0.25 [PITH_FULL_IMAGE:figures/full_fig_p020_10.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

33 extracted references · 19 canonical work pages · 1 internal anchor

  1. [1]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should not add it explicitly Type <Return> for now, but then later remove the command n...

  2. [2]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@first@sw \@firstoftwo \@ifundefined NAT@b*@#2 \@firstoftwo @num @NAT@ctr \@secondoft...

  3. [4]

    Trajectory-based air-writing recognition using deep neural network and depth sensor

    Md Shahinur Alam, Ki-Chul Kwon, Md Ashraful Alam, Mohammed Y Abbass, Shariar Md Imtiaz, and Nam Kim. Trajectory-based air-writing recognition using deep neural network and depth sensor. Sensors, 20 0 (2): 0 376, 2020

    work page 2020

  4. [5]

    Evolutionary clustering via message passing

    Natalia M Arzeno and Haris Vikalo. Evolutionary clustering via message passing. IEEE Transactions on Knowledge and Data Engineering, 33 0 (6): 0 2452--2466, 2019

    work page 2019

  5. [6]

    Dynamic local regret for non-convex online forecasting

    Sergul Aydore, Tianhao Zhu, and Dean P Foster. Dynamic local regret for non-convex online forecasting. Advances in neural information processing systems, 32, 2019

    work page 2019

  6. [7]

    Evolutionary clustering

    Deepayan Chakrabarti, Ravi Kumar, and Andrew Tomkins. Evolutionary clustering. In Proceedings of the 12th ACM SIGKDD international conference on Knowledge discovery and data mining, pages 554--560, 2006

    work page 2006

  7. [8]

    A simple framework for contrastive learning of visual representations

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. A simple framework for contrastive learning of visual representations. In International conference on machine learning, pages 1597--1607. PMLR, 2020

    2020

  8. [9]

    Exploring simple siamese representation learning

    Xinlei Chen and Kaiming He. Exploring simple siamese representation learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 15750--15758, 2021

    work page 2021

  9. [10]

    Fed-qssl: A framework for personalized federated learning under bitwidth and data heterogeneity

    Yiyue Chen, Haris Vikalo, and Chianing Wang. Fed-qssl: A framework for personalized federated learning under bitwidth and data heterogeneity. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 11443--11452, 2024

    work page 2024

  10. [11]

    Adaptive personalized federated learning

    Yuyang Deng, Mohammad Mahdi Kamani, and Mehrdad Mahdavi. Adaptive personalized federated learning. arXiv preprint arXiv:2003.13461, 2020

    Pith/arXiv arXiv 2003

  11. [12]

    Tslanet: Rethinking transformers for time series representation learning

    Emadeldeen Eldele, Mohamed Ragab, Zhenghua Chen, Min Wu, and Xiaoli Li. Tslanet: Rethinking transformers for time series representation learning. arXiv preprint arXiv:2404.08472, 2024

    Pith/arXiv arXiv 2024

  12. [13]

    SOM-VAE: Interpretable Discrete Representation Learning on Time Series

    Vincent Fortuin, Matthias H \"u ser, Francesco Locatello, Heiko Strathmann, and Gunnar R \"a tsch. Som-vae: Interpretable discrete representation learning on time series. arXiv preprint arXiv:1806.02199, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  13. [14]

    T-rep: Representation learning for time series using time-embeddings

    Archibald Fraikin, Adrien Bennetot, and St \'e phanie Allassonni \`e re. T-rep: Representation learning for time series using time-embeddings. arXiv preprint arXiv:2310.04486, 2023

    Pith/arXiv arXiv 2023

  14. [15]

    Unsupervised scalable representation learning for multivariate time series

    Jean-Yves Franceschi, Aymeric Dieuleveut, and Martin Jaggi. Unsupervised scalable representation learning for multivariate time series. Advances in neural information processing systems, 32, 2019

    work page 2019

  15. [16]

    An efficient framework for clustered federated learning

    Avishek Ghosh, Jichan Chung, Dong Yin, and Kannan Ramchandran. An efficient framework for clustered federated learning. Advances in Neural Information Processing Systems, 33: 0 19586--19597, 2020

    work page 2020

  16. [17]

    Long short-term memory

    S Hochreiter. Long short-term memory. Neural Computation MIT-Press, 1997

    work page 1997

  17. [18]

    How to escape saddle points efficiently

    Chi Jin, Rong Ge, Praneeth Netrapalli, Sham M Kakade, and Michael I Jordan. How to escape saddle points efficiently. In International conference on machine learning, pages 1724--1732. PMLR, 2017

    work page 2017

  18. [19]

    Time-llm: Time series forecasting by reprogramming large language models

    Ming Jin, Shiyu Wang, Lintao Ma, Zhixuan Chu, James Y Zhang, Xiaoming Shi, Pin-Yu Chen, Yuxuan Liang, Yuan-Fang Li, Shirui Pan, et al. Time-llm: Time series forecasting by reprogramming large language models. arXiv preprint arXiv:2310.01728, 2023

    Pith/arXiv arXiv 2023

  19. [20]

    Dynamic clustering in federated learning

    Yeongwoo Kim, Ezeddin Al Hakim, Johan Haraldson, Henrik Eriksson, Jos \'e Mairton B da Silva, and Carlo Fischione. Dynamic clustering in federated learning. In ICC 2021-IEEE International Conference on Communications, pages 1--6. IEEE, 2021

    work page 2021

  20. [21]

    Recent development and applications of sumo-simulation of urban mobility

    Daniel Krajzewicz, Jakob Erdmann, Michael Behrisch, and Laura Bieker. Recent development and applications of sumo-simulation of urban mobility. International journal on advances in systems and measurements, 5 0 (3&4), 2012

    work page 2012

  21. [22]

    Transfer learning promotes acquisition of individual bci skills

    Satyam Kumar, Hussein Alawieh, Frigyes Samuel Racz, Rawan Fakhreddine, and José del R Millán. Transfer learning promotes acquisition of individual bci skills. PNAS Nexus, 3 0 (2): 0 pgae076, 02 2024. ISSN 2752-6542. doi:10.1093/pnasnexus/pgae076. URL https://doi.org/10.1093/pnasnexus/pgae076

    doi:10.1093/pnasnexus/pgae076 2024

  22. [23]

    Federated learning with soft clustering

    Chengxi Li, Gang Li, and Pramod K Varshney. Federated learning with soft clustering. IEEE Internet of Things Journal, 9 0 (10): 0 7773--7782, 2021 a

    work page 2021

  23. [24]

    Federated optimization in heterogeneous networks

    Tian Li, Anit Kumar Sahu, Manzil Zaheer, Maziar Sanjabi, Ameet Talwalkar, and Virginia Smith. Federated optimization in heterogeneous networks. Proceedings of Machine learning and systems, 2: 0 429--450, 2020

    work page 2020

  24. [25]

    Ditto: Fair and robust federated learning through personalization

    Tian Li, Shengyuan Hu, Ahmad Beirami, and Virginia Smith. Ditto: Fair and robust federated learning through personalization. In International conference on machine learning, pages 6357--6368. PMLR, 2021 b

    work page 2021

  25. [26]

    Three approaches for personalization with applications to federated learning

    Yishay Mansour, Mehryar Mohri, Jae Ro, and Ananda Theertha Suresh. Three approaches for personalization with applications to federated learning. arXiv preprint arXiv:2002.10619, 2020

    Pith/arXiv arXiv 2002

  26. [27]

    Communication-efficient learning of deep networks from decentralized data

    Brendan McMahan, Eider Moore, Daniel Ramage, Seth Hampson, and Blaise Aguera y Arcas. Communication-efficient learning of deep networks from decentralized data. In Artificial intelligence and statistics, pages 1273--1282. PMLR, 2017

    2017

  27. [28]

    A greedy agglomerative framework for clustered federated learning

    Manan Mehta and Chenhui Shao. A greedy agglomerative framework for clustered federated learning. IEEE Transactions on Industrial Informatics, 19 0 (12): 0 11856--11867, 2023

    work page 2023

  28. [29]

    A time series is worth 64 words: Long-term forecasting with transformers

    Yuqi Nie, Nam H Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam. A time series is worth 64 words: Long-term forecasting with transformers. arXiv preprint arXiv:2211.14730, 2022

    Pith/arXiv arXiv 2022

  29. [30]

    Does learning from decentralized non-iid unlabeled data benefit from self supervision? arXiv preprint arXiv:2210.10947, 2022

    Lirui Wang, Kaiqing Zhang, Yunzhu Li, Yonglong Tian, and Russ Tedrake. Does learning from decentralized non-iid unlabeled data benefit from self supervision? arXiv preprint arXiv:2210.10947, 2022

    Pith/arXiv arXiv 2022

  30. [31]

    Timesnet: Temporal 2d-variation modeling for general time series analysis

    Haixu Wu, Tengge Hu, Yong Liu, Hang Zhou, Jianmin Wang, and Mingsheng Long. Timesnet: Temporal 2d-variation modeling for general time series analysis. arXiv preprint arXiv:2210.02186, 2022

    Pith/arXiv arXiv 2022

  31. [32]

    Adaptive evolutionary clustering

    Kevin S Xu, Mark Kliger, and Alfred O Hero III. Adaptive evolutionary clustering. Data Mining and Knowledge Discovery, 28: 0 304--336, 2014

    work page 2014

  32. [33]

    Metaclusterfl: Personalized federated learning on non-iid data with meta-learning and clustering

    Hui Zeng, Shiyu Xiong, and Hongzhou Shi. Metaclusterfl: Personalized federated learning on non-iid data with meta-learning and clustering. In 2024 International Joint Conference on Neural Networks (IJCNN), pages 1--10. IEEE, 2024

    work page 2024

  33. [34]

    Federated learning with non-iid data

    Yue Zhao, Meng Li, Liangzhen Lai, Naveen Suda, Damon Civin, and Vikas Chandra. Federated learning with non-iid data. arXiv preprint arXiv:1806.00582, 2018

    Pith/arXiv arXiv 2018

This paper was first reviewed by deepseek-v4-flash on August 4, 2026.