Pith. sign in

REVIEW 3 major objections 6 minor 63 references

Caesar: A Low-deviation Compression Approach for Efficient Federated Learning

T0 review · 3 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read Caesar cuts federated-learning traffic by up to 38% at same accuracy

desk verdict A solid FL systems paper with a clean combination of compression heuristics and real-hardware evaluation; the traffic-savings claim is plausible but needs error bars and sensitivity analysis on the importance model. read the letter →

arxiv 2412.19989 v1 pith:L2LGSYQB submitted 2024-12-28 cs.LG cs.DC

Jiaming Yan , Jianchun Liu , Hongli Xu , Liusheng Huang , Jiantao Gong , Xudong Liu , Kun Hou This is my paper · ORCID
classification cs.LGcs.DC
keywords federatedlearninggradientcompressionmodeldataheterogeneitystalenesscommunicationefficiencybatchsizeoptimizationdeviceimportance
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

Caesar is a federated learning framework built on a simple claim: the damage done by compression comes from model and gradient deviation, so compression ratios should be device-specific rather than fixed or capability-based. For each device, it sets the download compression ratio from the staleness of the device's local model and the upload ratio from the importance of the device's local data, then tunes batch sizes so faster devices do not wait idly. The paper reports that this reduces traffic by roughly 25.54% to 37.88% compared with compression-based baselines at the same target accuracy, while losing only 0.68% in final accuracy against full-precision communication. That matters because communication, not computation, is the usual bottleneck in mobile federated learning.

What carries the argument

The carrying mechanism is a pair of compression-ratio policies. For the downlink, the model compression ratio for device $i$ in round $t$ is $\rho_{M,i}^t=(1-\delta_i^t/t)\rho_M^{\max}$, where $\delta_i^t$ is the number of rounds since that device last participated; the server then applies Top-K sparsification plus 1-bit quantization, and the device recovers the 1-bit entries using its local model, replacing sign-mismatched or out-of-range values with the average absolute value. For the uplink, device importance is $C_i=\lambda n_i/n_{\max}+(1-\lambda)e^{-\mathrm{KL}(\Phi_i\|\Phi_0)}$ with $\lambda=0.5$ and a uniform global distribution $\Phi_0$, and a rank-based rule assigns smaller gradient compression ratios to higher-importance devices. A third mechanism, a greedy batch-size assignment, keeps each device's round time near the fastest device so the synchronous barrier does not waste time.

What would settle it

Run Caesar on a dataset with a deliberately skewed global label distribution, so the true $\Phi_0$ is far from uniform, and compare against the same system using the true distribution in the KL term; if the uniform-assumption version over-compresses high-contribution devices, its final accuracy or traffic savings should visibly degrade.

Watch

Extended reading notes

Core claim

The paper's central claim is that a low-deviation compression schedule can make federated learning nearly as accurate as uncompressed training while spending much less traffic. It argues that two sources of error dominate: devices with outdated local models need a faithful copy of the global model, and devices whose local gradients are most aligned with the global objective need faithful uploads. Caesar therefore gives stale devices smaller model compression ratios, gives data-rich and distributionally aligned devices smaller gradient compression ratios, and uses a greedy batch-size rule to equalize round durations. The measured consequence is that Caesar reaches a given target accuracy with 23.94% to 65.06% less traffic than four baselines across the tested datasets, keeping the final accuracy gap to full-precision communication at 0.68%.

Load-bearing premise

The load-bearing premise is that Eq. (5) correctly ranks devices by their contribution to global convergence using only sample volume and KL divergence from an assumed uniform global data distribution; if that ranking is wrong, important gradients get compressed too hard and the reported accuracy-traffic tradeoff weakens.

Editorial extensions

If this is right

  • At the target accuracy, federated deployments can spend roughly a quarter to a third less traffic than compression-based baselines, and up to about 65% less than individual baselines in the reported settings.
  • Frequently participating devices can accept high model compression ratios without hurting training, while rarely seen devices receive more faithful models, reducing bad updates from model obsolescence.
  • Devices with balanced label distributions and large sample volumes retain more gradient information, so the global model keeps learning the knowledge those devices carry.
  • Batch-size regulation shortens the synchronized round, so faster devices spend less time idle and the wall-clock time to target accuracy drops.
  • Because the compression schedule is method-agnostic, the same deviation-aware ratios can be applied on top of sparsification, quantization, or tensor-decomposition compressors.

Reading between the lines

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

  • If the global label distribution is not uniform, the fixed uniform $\Phi_0$ in the KL term could mis-rank devices; estimating it from the server's aggregated view or from recent participants would be a natural, testable correction.
  • The importance values are computed once before training; in federated settings where device populations or data distributions drift, periodic recomputation of importance may be needed to keep the rankings valid.
  • Because the savings come from ratios applied to whole models, the method should benefit larger models even more; testing on a modern transformer-scale model would show whether the trend holds.
  • The importance ranking could also be reused by a participant-selection strategy, choosing the most important devices more often would likely reduce rounds further, though this is not what the paper itself claims.
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

3 major / 6 minor

Summary. Caesar is a federated learning framework that adaptively compresses both downlink global-model updates and uplink local gradients. The server sets each device's model compression ratio from its participation staleness (Eq. 3), compresses via a hybrid Top-K sparsification plus 1-bit quantization scheme with local-model-based recovery, and computes a device importance score from sample volume and KL divergence to a uniform reference distribution (Eqs. 4-5), converting importance ranks into gradient compression ratios (Eq. 6). A greedy batch-size rule (Eq. 9) attempts to equalize per-round completion times. Experiments on 80 NVIDIA Jetson devices and 40 OPPO smartphones across four datasets report 23.94%-65.06% traffic reductions relative to FedAvg, FlexCom, ProWD, and PyramidFL at a common target accuracy, with only a 0.68% final-accuracy loss relative to full-precision communication.

Significance. The core idea of Caesar—tying compression fidelity to staleness and data quality rather than only to device capability—is timely and likely useful for cross-device FL systems. The paper's strengths include physical implementation on two hardware platforms, a wrapper design that is method-agnostic to the underlying gradient compression operator, and the absence of fitted constants for the headline traffic claims: λ=0.5 is fixed by default and the compression bounds [0.1, 0.6] are adopted from prior work. If the reported traffic reductions reproduce under multi-seed evaluation and under a data-dependent global-distribution estimate, this would be a solid systems contribution. The current evidence is limited by the evaluation methodology and by an unsupported uniform-global-distribution assumption in the importance model.

major comments (3)
  1. [§4.2, Eqs. (4)-(5)] The assertion that "the global data distribution is typically uniform" (p_h^0 = 1/K) is load-bearing for Caesar's gradient compression, and it is contradicted by the paper's own evaluation settings. In the Dirichlet-based splits (§6.1), the aggregate realized data distribution is only approximately uniform for large N, and in OPPO-TS the data are assigned by real user IDs with no reason to be uniform. Under a non-uniform global distribution, a device with a large volume of rare-class data can receive a low importance score because KL(Φ_i || uniform) is large, and Eq. (6) then assigns it a high compression ratio. Since Table 3 measures traffic at a fixed target accuracy, any accuracy loss from over-compressing such gradients can force extra rounds and erode the claimed 25.54%-37.88% traffic reduction. I request a sensitivity analysis over λ and over the reference distribution Φ0 (e.g., a server-side estimate from reported label frequencies), and at least one experiment with a deliberately skewed global class distribution.
  2. [Table 3 and §6.1-§6.2] The evaluation reports single runs with no random seeds, confidence intervals, or error bars. The target accuracy/AUC in Table 3 is described as "the highest achievable by all schemes," which appears to be chosen after observing the runs; comparing at a post hoc ceiling can favor the scheme that degrades least at that point and can misstate the traffic-to-accuracy ratio. Please prespecify target accuracies before running, report results across at least three seeds with variance, and show traffic versus a range of target accuracies rather than a single point. Without this, the precise percentages in the abstract and Table 3 are not yet convincing.
  3. [§6.4] The ablation study does not cleanly support the stated contributions. Caesar takes 8,615s and 115.57GB to reach 80% on CIFAR-10; Caesar-BR (no deviation-aware compression) takes 12,043s and 228.31GB; Caesar-DC (no batch-size regulation) takes 17,818s and 147.59GB. The text says adaptive batch-size regulation gives a 2.09× speedup, but 17,818/8,615 ≈ 2.07, and says deviation-aware compression gives a 1.39× speedup, but 12,043/8,615 ≈ 1.40; the traffic-reduction contributions (21.69% vs 49.38%) are computed against different baselines. The conclusion that deviation-aware compression "plays a more critical role" is therefore not cleanly established. Please provide a full 2×2 ablation and report all ratios against the same full-Caesar baseline.
minor comments (6)
  1. [§3] Typo: "gradiens" should be "gradients" in the overview sentence.
  2. [§4.3] Typo: "dose not exceeds" should be "does not exceed" in the explanation following Eq. (9).
  3. [§4.1] The number of staleness clusters E is a tunable complexity-accuracy knob but is never specified in the experiments; please report the value used or a sensitivity study.
  4. [Figure 1] The axis label and legend in Figure 1(b) are partly garbled in the PDF rendering; please regenerate the figure.
  5. [§7] Minor capitalization issue: "Meanwhile, We impose" should be "Meanwhile, we impose."
  6. [§5] The manuscript does not mention code release; given the 3k-line implementation, providing an artifact would substantially strengthen reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: Caesar's compression ratios are set by closed-form heuristics from device properties and are evaluated against external baselines, not fitted to the headline traffic numbers.

full rationale

I walked Caesar's derivation chain and found no step where a claimed prediction reduces to its own inputs. The global model compression ratio is defined by Eq. (3) as a deterministic function of staleness: r_{m,i}^t = (1 - s_i^t / t) * r_max^m. The local gradient compression ratio is defined by Eq. (6) from a rank of the importance score C_i, which is itself computed by Eq. (5) from sample volume and KL divergence to a uniform reference distribution. These formulas are design heuristics; they are not fitted to the measured traffic-to-accuracy outcomes, so the reported 25.54%-37.88% traffic reductions are empirical measurements at a fixed target accuracy, not constructions. The rank-based strategy is explicitly attributed to external PyramidFL [36], and the compression bounds [0.1, 0.6] are also taken from [36], so no load-bearing self-citation chain appears. The claim that the global data distribution 'is typically uniform' in Eq. (4)'s surrounding text is asserted without support and is a modeling weakness, but it is not circular: even if the assumption is wrong, the algorithm's outputs would be suboptimal or inaccurate, not identical to its inputs by definition. The evaluation compares actual measured traffic and training time against four external baselines on physical testbeds, with target accuracies chosen as the highest achievable by all schemes; this provides an independent check of the claimed advantage. No constants are fitted to the target accuracy, and no quantity called a prediction is actually recovered from the data being predicted. Thus the appropriate finding is no significant circularity.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The paper introduces two heuristic formulas (importance C_i and compression ratio schedules) but no new physical entities. The central claim rests on the uniform-global-distribution assumption, the ad hoc importance metric, and the linear recovery-time model.

free parameters (3)
  • λ (importance weight) = 0.5
    Eq. (5) balances normalized sample volume and inverse exponential KL divergence. Set by default with no sensitivity analysis; directly controls gradient compression ratios.
  • Compression ratio bounds θ_max_m, θ_min_g, θ_max_g = 0.1 to 0.6
    Taken from PyramidFL [36] and treated as system hyperparameters; not fitted to the target accuracy, but they constrain all compression ratios in the evaluation.
  • Maximum batch size B_max = not stated explicitly
    Predefined in Eq. (8); used to compute the fastest device anchor for batch size equalization.
assumptions (5)
  • domain assumption The global data distribution is uniform (Φ0 = 1/K), so KL divergence to uniform measures gradient importance.
    Eq. (4) defines the distribution gap using uniform Φ0. In real non-IID FL the global distribution is a weighted mixture of client distributions and is generally not uniform; a non-uniform Φ0 could change the importance ranking.
  • ad hoc to paper Device importance is a weighted sum of normalized sample volume and inverse exponential KL divergence (Eq. 5).
    No derivation or empirical validation that this scalar predicts a device's contribution to global convergence; it is the core heuristic for gradient compression ratios.
  • domain assumption A device with staleness s can recover a compressed global model by replacing 1-bit parameters with its local model values, and recovery quality scales linearly with (1 - s/t).
    Section 4.1: the recovery mechanism assumes local model elements approximate the omitted global elements; Eq. (3) assumes a linear relationship between staleness and needed compression ratio.
  • domain assumption Per-round device time is T_i = θ_m·(S/B_m) + θ_g·(S/B_g) + τ·b_i·μ_i (Eq. 7).
    A linear additive model for download, upload, and compute; ignores queuing, contention, and bandwidth dynamics, yet is used to set batch sizes.
  • standard math Top-K sparsification plus 1-bit quantization preserves enough information for FL convergence when ratios are guided by importance.
    Borrowed from prior compression literature; the paper does not prove convergence for the adaptive scheme, only relies on the empirical performance of Top-K and quantization.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Caesar: A Low-deviation Compression Approach for Efficient Federated Learning." pith.science (2026). https://pith.science/paper/L2LGSYQB

@misc{pith2026241219989,
  author       = {Pith},
  title        = {Pith review of: Caesar: A Low-deviation Compression Approach for Efficient Federated Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L2LGSYQB}},
  note         = {Machine review of arXiv:2412.19989}
}
abstract

Compression is an efficient way to relieve the tremendous communication overhead of federated learning (FL) systems. However, for the existing works, the information loss under compression will lead to unexpected model/gradient deviation for the FL training, significantly degrading the training performance, especially under the challenges of data heterogeneity and model obsolescence. To strike a delicate trade-off between model accuracy and traffic cost, we propose Caesar, a novel FL framework with a low-deviation compression approach. For the global model download, we design a greedy method to optimize the compression ratio for each device based on the staleness of the local model, ensuring a precise initial model for local training. Regarding the local gradient upload, we utilize the device's local data properties (\ie, sample volume and label distribution) to quantify its local gradient's importance, which then guides the determination of the gradient compression ratio. Besides, with the fine-grained batch size optimization, Caesar can significantly diminish the devices' idle waiting time under the synchronized barrier. We have implemented Caesar on two physical platforms with 40 smartphones and 80 NVIDIA Jetson devices. Extensive results show that Caesar can reduce the traffic costs by about 25.54%$\thicksim$37.88% compared to the compression-based baselines with the same target accuracy, while incurring only a 0.68% degradation in final test accuracy relative to the full-precision communication.

Figures

Figures reproduced from arXiv: 2412.19989 by the authors.

Figure 1
Figure 1. The results of preliminary experiments. (a) The training process of different FL approaches on CIFAR-10 with 250 communication rounds. (b) The traffic costs of different FL approaches to achieve a target accuracy of 72% on CIFAR-10; (c) The relationship between initial model error, model compression ratio, and local model staleness; (d) The importance of devices’ local gradients and the adopted gradient compression … view at source ↗
Figure 2
Figure 2. Overview of Caesar’s workflow. local models’ staleness, which guides the determination of the global model compression ratio ( 1 ). After that, the PS dispatches the compressed global models to corresponding participants ( 2 ). Once receiving the model, each device re￾covers the original global model via its local model. Since the model/gradient compression ratios vary across devices, their transmission latencies di… view at source ↗
Figure 3
Figure 3. Illustration of the model recovery mechanism in Caesar with a compression ratio of 5/9. subsequent model recovery. As illustrated in [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: The system architecture of Caesar. 7 [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Time-to-Accuracy performance of five schemes on the four datasets. 0.70 0.75 0.80 0 6 12 18 24 Traffic Cost (×10GB) Target Accuracy Caesar ProWD FedAvg PyramidFL FlexCom (a) CIFAR-10 0.80 0.83 0.86 0 1 2 3 4 5 Traffic Cost (GB) Target Accuracy Caesar ProWD FedAvg Pyram…
Figure 6
Figure 6. Figure 6: Traffic-to-Accuracy performance of five schemes on the four datasets. time cost. For instance, by [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: The average waiting time among participants of five schemes on the four datasets. 1 2 4 5 10 0.45 0.60 0.75 0.90 Test Accuracy Data Heterogeneity Level Caesar ProWD PyramidFL FlexCom FedAvg (a) CIFAR-10 1 2 4 5 10 0.72 0.76 0.80 0.84 0.88 Test Accuracy Data Heterogenei…
Figure 8
Figure 8. Figure 8: The effect of different data heterogeneity levels on five schemes’ training performance. the training efficiency. Therefore, Caesar can enjoy better traffic/time-to-accuracy performance than the baselines. To further demonstrate Caesar’s advantage under the syn￾chronou…
Figure 9
Figure 9. Figure 9: Caesar’s ablation study. shown in [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 10
Figure 10. Figure 10: Performance under different device scales. 8 Conclusion In this work, we propose Caesar, an efficient FL system with a novel deviation-aware compression approach. Specifically, to address the challenge of model obsolescence,Caesar cherry￾picks the downstream compressi…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 58 canonical work pages

  1. [1]

    Billion-scale feder- ated learning on mobile clients: A submodel design with tunable pri- vacy

    Chaoyue Niu, Fan Wu, Shaojie Tang, Lifeng Hua, Rongfei Jia, Chengfei Lv, Zhihua Wu, and Guihai Chen. Billion-scale feder- ated learning on mobile clients: A submodel design with tunable pri- vacy. In Proceedings of the 26th Annual International Conference on Mobile Computing and Networking, pages 1–14, 2020

    work page 2020

  2. [2]

    Aut- ofed: Heterogeneity-aware federated multimodal learning for rob ust autonomous driving

    Tianyue Zheng, Ang Li, Zhe Chen, Hongbo Wang, and Jun Luo. Aut- ofed: Heterogeneity-aware federated multimodal learning for rob ust autonomous driving. In Proceedings of the 29th Annual International Conference on Mobile Computing and Networking, pages 1–15, 2023

    work page 2023

  3. [3]

    General data protection regulation — Wikipedia, the free encyclopedia, 2024

    Wikipedia contributors. General data protection regulation — Wikipedia, the free encyclopedia, 2024. [Online; accessed 20- February-2024]

    work page 2024

  4. [4]

    California consumer privacy act — Wikipedia, the free encyclopedia, 2024

    Wikipedia contributors. California consumer privacy act — Wikipedia, the free encyclopedia, 2024. [Online; accessed 20- February-2024]

    work page 2024

  5. [5]

    Communication-efficient learning of deep net- works from decentralized data

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

    work page 2017

  6. [6]

    Yoga: Adaptive layer-wise model aggregation for decen- tralized federated learning

    Jun Liu, Jianchun Liu, Hongli Xu, Yunming Liao, Zhiyuan Wang, and Qianpiao Ma. Yoga: Adaptive layer-wise model aggregation for decen- tralized federated learning. IEEE/ACM Transactions on Networking, 2023

    work page 2023

  7. [7]

    Adaptive block-wise regularization and knowl- edge distillation for enhancing federated learning

    Jianchun Liu, Qingmin Zeng, Hongli Xu, Yang Xu, Zhiyuan Wang, and He Huang. Adaptive block-wise regularization and knowl- edge distillation for enhancing federated learning. IEEE/ACM Transactions on Networking, 2023

    work page 2023

  8. [8]

    Adaptive asynchronous federated learning in resource-constrained edge computing

    Jianchun Liu, Hongli Xu, Lun Wang, Yang Xu, Chen Qian, Jinyang Huang, and He Huang. Adaptive asynchronous federated learning in resource-constrained edge computing. IEEE Transactions on Mobile Computing, 22(2):674–690, 2021

    work page 2021

Show all 63 references
  1. [9]

    Fedmp: Federated learning through adaptive model pruning in heterogeneous edge computing

    Zhida Jiang, Yang Xu, Hongli Xu, Zhiyuan Wang, Chunming Qiao, and Yangming Zhao. Fedmp: Federated learning through adaptive model pruning in heterogeneous edge computing. In 2022 IEEE 38th International Conference on Data Engineering (ICDE), pages 767–779. IEEE, 2022

    2022

  2. [10]

    Bert: Pre-training of deep bidirectional transformers for language un- derstanding

    Jacob Devlin Ming-Wei Chang Kenton and Lee Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language un- derstanding. In Proceedings of naacL-HLT, volume 1, page 2. Min- neapolis, Minnesota, 2019

    2019

  3. [11]

    Newsweeder: Learning to filter netnews

    Ken Lang. Newsweeder: Learning to filter netnews. In Machine learning proceedings 1995, pages 331–339. Elsevier, 1995

    1995

  4. [12]

    Fednlp: Benchmarking federated learn- ing methods for natural language processing tasks

    Bill Yuchen Lin, Chaoyang He, Zihang Zeng, Hulin Wang, Yufen Huang, Christophe Dupuy, Rahul Gupta, Mahdi Soltanolkotabi, Xiang Ren, and Salman Avestimehr. Fednlp: Benchmarking federated learn- ing methods for natural language processing tasks. arXiv preprint arXiv:2104.08815, 2021

    2021 arXiv

  5. [13]

    Efficient federated learning for modern nlp

    Dongqi Cai, Yaozong Wu, Shangguang Wang, Felix Xiaozhu Lin, and Mengwei Xu. Efficient federated learning for modern nlp. In Proceedings of the 29th Annual International Conference on Mobile Computing and Networking, pages 1–16, 2023

    2023

  6. [14]

    Federated learning meets blockchain in edge computing: Oppor- tunities and challenges

    Dinh C Nguyen, Ming Ding, Quoc-Viet Pham, Pubudu N Pathirana, Long Bao Le, Aruna Seneviratne, Jun Li, Dusit Niyato, and H Vincent Poor. Federated learning meets blockchain in edge computing: Oppor- tunities and challenges. IEEE Internet of Things Journal , 8(16):12806– 12825, 2021

    2021

  7. [15]

    Advances and open problems in federated learning

    Peter Kairouz, H Brendan McMahan, Brendan Avent, Aurélien Bel - let, Mehdi Bennis, Arjun Nitin Bhagoji, Kallista Bonawitz, Zachary Charles, Graham Cormode, Rachel Cummings, et al. Advances and open problems in federated learning. Foundations and trends® in machine learning, 14...

    2021

  8. [16]

    Heterogeneity-aware federated learning with adaptive client selec- tion and gradient compression

    Zhida Jiang, Yang Xu, Hongli Xu, Zhiyuan Wang, and Chen Qian. Heterogeneity-aware federated learning with adaptive client selec- tion and gradient compression. In IEEE INFOCOM 2023-IEEE Conference on Computer Communications, pages 1–10. IEEE, 2023

    2023

  9. [17]

    Accelerating decentralized federated learning in heterogeneous edg e computing

    Lun Wang, Yang Xu, Hongli Xu, Min Chen, and Liusheng Huang. Accelerating decentralized federated learning in heterogeneous edg e computing. IEEE Transactions on Mobile Computing, 22(9):5001– 5016, 2023

    2023

  10. [18]

    Optimizing federated learning on non-iid data with reinforcement learning

    Hao Wang, Zakhary Kaplan, Di Niu, and Baochun Li. Optimizing federated learning on non-iid data with reinforcement learning. In IEEE INFOCOM 2020-IEEE conference on computer communications , pages 1698–1707. IEEE, 2020

    2020

  11. [19]

    Enhancing federated learning with intelligent model migration in heterogeneous edge computing

    Jianchun Liu, Yang Xu, Hongli Xu, Yunming Liao, Zhiyuan Wang, and He Huang. Enhancing federated learning with intelligent model migration in heterogeneous edge computing. In 2022 IEEE 38th International Conference on Data Engineering (ICDE), pages 1586–

    2022

  12. [20]

    Towards federated learning at scale: System design

    Keith Bonawitz, Hubert Eichner, Wolfgang Grieskamp, Dzmitry Huba, Alex Ingerman, Vladimir Ivanov, Chloe Kiddon, Jakub Konečn `y, Ste- fano Mazzocchi, Brendan McMahan, et al. Towards federated learning at scale: System design. Proceedings of machine learning and systems, 1:374–...

    2019

  13. [21]

    Op- timal rate adaption in federated learning with compressed commu- nications

    Laizhong Cui, Xiaoxin Su, Yipeng Zhou, and Jiangchuan Liu. Op- timal rate adaption in federated learning with compressed commu- nications. In IEEE INFOCOM 2022-IEEE Conference on Computer Communications, pages 1459–1468. IEEE, 2022

    2022

  14. [22]

    Deepreduce: A sparse-tensor communi- cation framework for federated deep learning

    Hang Xu, Kelly Kostopoulou, Aritra Dutta, Xin Li, Alexandros Ntoulas, and Panos Kalnis. Deepreduce: A sparse-tensor communi- cation framework for federated deep learning. Advances in Neural 13 Information Processing Systems, 34:21150–21163, 2021

    2021

  15. [23]

    Ron Dorfman, Shay Vargaftik, Yaniv Ben-Itzhak, and Kfir Y. Levy . Docofl: downlink compression for cross-device federated learning. In Proceedings of the 40th International Conference on Machine Learning, ICML’23. JMLR.org, 2023

    2023

  16. [24]

    signsgd: Compressed optimisation for non- convex problems

    Jeremy Bernstein, Yu-Xiang Wang, Kamyar Azizzadenesheli, and An- imashree Anandkumar. signsgd: Compressed optimisation for non- convex problems. In International Conference on Machine Learning, pages 560–569. PMLR, 2018

    2018

  17. [25]

    To talk or to work: Flexible communication compression for en- ergy efficient federated learning over heterogeneous mobile edge devices

    Liang Li, Dian Shi, Ronghui Hou, Hui Li, Miao Pan, and Zhu Han. To talk or to work: Flexible communication compression for en- ergy efficient federated learning over heterogeneous mobile edge devices. In IEEE INFOCOM 2021-IEEE Conference on Computer Communications, pages 1–10. I...

    2021

  18. [26]

    Adaptive control of local updating and model compres- sion for efficient federated learning

    Yang Xu, Yunming Liao, Hongli Xu, Zhenguo Ma, Lun Wang, and Jianchun Liu. Adaptive control of local updating and model compres- sion for efficient federated learning. IEEE Transactions on Mobile Computing, 2022

    2022

  19. [27]

    Resource- adaptive federated learning with all-in-one neural composition

    Yiqun Mei, Pengfei Guo, Mo Zhou, and Vishal Patel. Resource- adaptive federated learning with all-in-one neural composition. Advances in Neural Information Processing Systems, 35:4270–4284, 2022

    2022

  20. [28]

    Communication-efficient federated learning for heterogeneous edge devices based on adapt ive gradient quantization

    Heting Liu, Fang He, and Guohong Cao. Communication-efficient federated learning for heterogeneous edge devices based on adapt ive gradient quantization. In IEEE INFOCOM 2023-IEEE Conference on Computer Communications, pages 1–10. IEEE, 2023

    2023

  21. [29]

    Peaches: Personalized federated learning with neural ar- chitecture search in edge computing

    Jiaming Yan, Jianchun Liu, Hongli Xu, Zhiyuan Wang, and Chun- ming Qiao. Peaches: Personalized federated learning with neural ar- chitecture search in edge computing. IEEE Transactions on Mobile Computing, 2024

    2024

  22. [30]

    Feddisco: Federated learning with discrepancy-aware col- laboration

    Rui Ye, Mingkai Xu, Jianyu Wang, Chenxin Xu, Siheng Chen, and Yan- feng Wang. Feddisco: Federated learning with discrepancy-aware col- laboration. arXiv preprint arXiv:2305.19229, 2023

    2023 arXiv

  23. [31]

    Bose: Block-wise federated learning in hetero- geneous edge computing

    Lun Wang, Yang Xu, Hongli Xu, Zhida Jiang, Min Chen, Wuyang Zhang, and Chen Qian. Bose: Block-wise federated learning in hetero- geneous edge computing. IEEE/ACM Transactions on Networking, 2023

    2023

  24. [32]

    Efficient mini-batch training for stochastic optimization

    Mu Li, Tong Zhang, Yuqiang Chen, and Alexander J Smola. Efficient mini-batch training for stochastic optimization. In Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining, pages 661–670, 2014

    2014

  25. [33]

    Measuring the effects of non-identical data distribution for federated visual class ifi- cation

    Tzu-Ming Harry Hsu, Hang Qi, and Matthew Brown. Measuring the effects of non-identical data distribution for federated visual class ifi- cation. arXiv preprint arXiv:1909.06335, 2019

    1909 arXiv

  26. [34]

    Bayesian nonparamet- ric federated learning of neural networks

    Mikhail Yurochkin, Mayank Agarwal, Soumya Ghosh, Kristjan Gr ee- newald, Nghia Hoang, and Yasaman Khazaeni. Bayesian nonparamet- ric federated learning of neural networks. In International conference on machine learning, pages 7252–7261. PMLR, 2019

    2019

  27. [35]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 770– 778, 2016

    2016

  28. [36]

    Pyramidfl: A fine-grained client selection framework for efficient federated learn- ing

    Chenning Li, Xiao Zeng, Mi Zhang, and Zhichao Cao. Pyramidfl: A fine-grained client selection framework for efficient federated learn- ing. In Proceedings of the 28th Annual International Conference on Mobile Computing And Networking, pages 158–171, 2022

    2022

  29. [37]

    Oort: Efficient federated learning via guided participant selec- tion

    Fan Lai, Xiangfeng Zhu, Harsha V Madhyastha, and Mosharaf Chowd- hury. Oort: Efficient federated learning via guided participant selec- tion. In 15th{USENIX} Symposium on Operating Systems Design and Implementation ({OSDI} 21), pages 19–35, 2021

    2021

  30. [38]

    The convergence of sparsi- fied gradient methods

    Dan Alistarh, Torsten Hoefler, Mikael Johansson, Nikola Konst anti- nov, Sarit Khirirat, and Cédric Renggli. The convergence of sparsi- fied gradient methods. Advances in Neural Information Processing Systems, 31, 2018

    2018

  31. [39]

    Mergesfl: Split federated learning with feature merging and batch size regulation

    Yunming Liao, Yang Xu, Hongli Xu, Lun Wang, Zhiwei Yao, and Chun- ming Qiao. Mergesfl: Split federated learning with feature merging and batch size regulation. In 2024 IEEE 40th International Conference on Data Engineering (ICDE), pages 2054–2067. IEEE, 2024

    2024

  32. [40]

    Approximating the kull- back leibler divergence between gaussian mixture models

    John R Hershey and Peder A Olsen. Approximating the kull- back leibler divergence between gaussian mixture models. In 2007 IEEE International Conference on Acoustics, Speech and Signal Processing-ICASSP’07, volume 4, pages IV–317. IEEE, 2007

    2007

  33. [41]

    An efficient image similarity measure based on approximations of kl-divergence between two gaussian mixtures

    Goldberger and Greenspan. An efficient image similarity measure based on approximations of kl-divergence between two gaussian mixtures. In Proceedings Ninth IEEE International conference on computer vision, pages 487–493. IEEE, 2003

    2003

  34. [42]

    Adaptive batch size for federated learning in resource-constrained edge computing

    Zhenguo Ma, Yang Xu, Hongli Xu, Zeyu Meng, Liusheng Huang, and Yinxing Xue. Adaptive batch size for federated learning in resource-constrained edge computing. IEEE Transactions on Mobile Computing, 22(1):37–53, 2023

    2023

  35. [43]

    Pytorch: An imperative style, high- performance deep learning library

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, Jame s Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. Pytorch: An imperative style, high- performance deep learning library. Advances in neural information processing systems, 32, 2019

    2019

  36. [44]

    Mnn: A universal and efficient inference engine

    Xiaotang Jiang, Huan Wang, Yiliu Chen, Ziqi Wu, Lichuan Wang, Bin Zou, Yafeng Yang, Zongyang Cui, Yu Cai, Tianhang Yu, et al. Mnn: A universal and efficient inference engine. Proceedings of Machine Learning and Systems, 2:1–13, 2020

    2020

  37. [45]

    Building a virtual system of systems using docker swar m in multiple clouds

    Nitin Naik. Building a virtual system of systems using docker swar m in multiple clouds. In 2016 IEEE International Symposium on Systems Engineering (ISSE), pages 1–3. IEEE, 2016

    2016

  38. [46]

    mpi4py: Status update afte r 12 years of development

    Lisandro Dalcin and Yao-Lung L Fang. mpi4py: Status update afte r 12 years of development. Computing in Science & Engineering, 23(4):47– 54, 2021

    2021

  39. [47]

    Array programming with numpy

    Charles R Harris, K Jarrod Millman, Stéfan J Van Der Walt, Ra lf Gom- mers, Pauli Virtanen, David Cournapeau, Eric Wieser, Julian Taylor, Sebastian Berg, Nathaniel J Smith, et al. Array programming with numpy. Nature, 585(7825):357–362, 2020

    2020

  40. [48]

    Learning multiple layer s of features from tiny images

    Alex Krizhevsky, Geoffrey Hinton, et al. Learning multiple layer s of features from tiny images. 2009

    2009

  41. [49]

    A public domain dataset for human a c- tivity recognition using smartphones

    Davide Anguita, Alessandro Ghio, Luca Oneto, Xavier Parra, Jorge Luis Reyes-Ortiz, et al. A public domain dataset for human a c- tivity recognition using smartphones. In Esann, volume 3, page 3, 2013

    2013

  42. [50]

    Speech commands: A dataset for limited-vocabu lary speech recognition

    Pete Warden. Speech commands: A dataset for limited-vocabu lary speech recognition. arXiv preprint arXiv:1804.03209, 2018

    2018 arXiv

  43. [51]

    Bitwidth heterogeneous federated learning with progressive weight dequantization

    Jaehong Yoon, Geon Park, Wonyong Jeong, and Sung Ju Hwang. Bitwidth heterogeneous federated learning with progressive weight dequantization. In International Conference on Machine Learning, pages 25552–25565. PMLR, 2022

    2022

  44. [52]

    Towards efficient communications in fed- erated learning: A contemporary survey

    Zihao Zhao, Yuzhu Mao, Yang Liu, Linqi Song, Ye Ouyang, Xinlei Chen, and Wenbo Ding. Towards efficient communications in fed- erated learning: A contemporary survey. Journal of the Franklin Institute, 360(12):8669–8703, 2023

    2023

  45. [53]

    Dadaquant: Doubl y- adaptive quantization for communication-efficient federated learning

    Robert Hönig, Yiren Zhao, and Robert Mullins. Dadaquant: Doubl y- adaptive quantization for communication-efficient federated learning. In International Conference on Machine Learning, pages 8852–8866. PMLR, 2022

    2022

  46. [54]

    Qsgd: Communication-efficient sgd via gradient quantization and encoding

    Dan Alistarh, Demjan Grubic, Jerry Li, Ryota Tomioka, and Mila n Vo- jnovic. Qsgd: Communication-efficient sgd via gradient quantization and encoding. Advances in neural information processing systems, 30, 2017

    2017

  47. [55]

    Robust and communication-efficient federated learning from non-iid data

    Felix Sattler, Simon Wiedemann, Klaus-Robert Müller, and Woj ciech Samek. Robust and communication-efficient federated learning from non-iid data. IEEE transactions on neural networks and learning systems, 31(9):3400–3413, 2019. 14

    2019

  48. [56]

    Fjord: Fair and accurate fed- erated learning under heterogeneous targets with ordered dropout

    Samuel Horvath, Stefanos Laskaridis, Mario Almeida, Ilias Le ontiadis, Stylianos Venieris, and Nicholas Lane. Fjord: Fair and accurate fed- erated learning under heterogeneous targets with ordered dropout . Advances in Neural Information Processing Systems, 34:12876–12889, 2021

    2021

  49. [57]

    Heterofl: Computation and communication efficient federated learning for heterogeneous clients

    Enmao Diao, Jie Ding, and Vahid Tarokh. Heterofl: Computation and communication efficient federated learning for heterogeneous clients. In International Conference on Learning Representations, 2020

    2020

  50. [58]

    Svdfed: Enabling communication-efficient federated learning via singular- value-decomposition

    Haolin Wang, Xuefeng Liu, Jianwei Niu, and Shaojie Tang. Svdfed: Enabling communication-efficient federated learning via singular- value-decomposition. In IEEE INFOCOM 2023-IEEE Conference on Computer Communications, pages 1–10. IEEE, 2023

    2023

  51. [59]

    Fedpara: Low- rank hadamard product for communication-efficient federated learn- ing

    Nam Hyeon-Woo, Moon Ye-Bin, and Tae-Hyun Oh. Fedpara: Low- rank hadamard product for communication-efficient federated learn- ing. In International Conference on Learning Representations, 2021

    2021

  52. [60]

    Hermes: an efficient federated learning framework for heterogeneous mobile clients

    Ang Li, Jingwei Sun, Pengcheng Li, Yu Pu, Hai Li, and Yiran Chen. Hermes: an efficient federated learning framework for heterogeneous mobile clients. In Proceedings of the 27th Annual International Conference on Mobile Computing and Networking , pages 420–437, 2021

    2021

  53. [61]

    When edge meets learning: Adaptive control for resource-constrained distributed ma- chine learning

    Shiqiang Wang, Tiffany Tuor, Theodoros Salonidis, Kin K Leung, Christian Makaya, Ting He, and Kevin Chan. When edge meets learning: Adaptive control for resource-constrained distributed ma- chine learning. In IEEE INFOCOM 2018-IEEE conference on computer communications, pages 6...

    2018

  54. [62]

    Tackling the objective inconsistency problem in heterogeneous federated optimization

    Jianyu Wang, Qinghua Liu, Hao Liang, Gauri Joshi, and H Vincent Poor. Tackling the objective inconsistency problem in heterogeneous federated optimization. Advances in neural information processing systems, 33:7611–7623, 2020

    2020

  55. [63]

    Tackling system and statistical heterogeneity for feder ated learning with adaptive client sampling

    Bing Luo, Wenli Xiao, Shiqiang Wang, Jianwei Huang, and Leandros Tassiulas. Tackling system and statistical heterogeneity for feder ated learning with adaptive client sampling. In IEEE INFOCOM 2022-IEEE conference on computer communications, pages 1739–1748. IEEE, 2022. 15

    2022

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.