Pith. sign in

REVIEW 4 major objections 5 minor 56 references

FedA2L: Adaptive layer-wise learning rate adjustment in decentralized federated learning

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read FedA2L claims that assigning each layer its own learning rate, derived from locally observable divergence signals, reliably accelerates convergence in decentralized federated learning — up to 4.94x faster than vanilla DFL and 59% fewer…

desk verdict Useful plug-in LR adaptation for DFL with a clean empirical sweep, but the layer-wise attribution is untested and hyperparameters are tuned on the evaluation benchmark. read the letter →

arxiv 2608.09208 v1 pith:TEFTEPFW submitted 2026-08-10 cs.LG cs.AIcs.CVcs.DC

Van Truong Vo , Khoa Nguyen , Taehong Kim This is my paper · ORCID
classification cs.LGcs.AIcs.CVcs.DC
keywords DecentralizedfederatedlearningLayer-wiserateadaptationDataheterogeneityNetworkconsensusConvergenceaccelerationCommunicationefficiencyInternetofThings(IoT)EdgeComputing
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

This paper tries to establish that a uniform learning rate across all neural-network layers is a primary brake on convergence when data is distributed unevenly across nodes in decentralized federated learning. It proposes FedA2L, which gives every layer its own learning rate each round, computed from two signals that are already present in the local training loop: how much the layer's weights moved during local training, and how stable the layer remained when neighbors' models were averaged in. FedA2L claims to reach target accuracy up to 4.94x faster than vanilla DFL and to use up to 59% fewer communication rounds than scheduler-based baselines, across six DFL algorithms and several architectures and datasets. The method adds no communication overhead and no central coordination, and the paper provides convergence theorems showing the usual decentralized-SGD rates are preserved. If the claim holds, a cheap per-layer rate adjustment could make large-scale edge and IoT learning markedly less communication-hungry.

What carries the argument

The central object is the three-state decomposition of a DFL round: base state $\theta^{B,r}_{i,l}$, trained state $\theta^{T,r}_{i,l}$, and aggregated state $\theta^{A,r}_{i,l}$, with the aggregated state becoming the next round's base state. From these, FedA2L computes two per-layer metrics — weight divergence $\sigma = \|\theta^T - \theta^B\|_2 / (\|\theta^B\|_2 + \epsilon)$ and aggregation stability $\zeta$, the fraction of parameters whose change under aggregation is below a threshold $\tau$ — then normalizes them with Z-scores over a window of $\rho$ rounds. The normalized signals are combined into a fusion score $\lambda = \beta e^{\omega} + (1-\beta) e^{\delta}$, which is mapped to a learning-rate multiplier via $1 + \tanh(\log \lambda)$ with global decay factor $\gamma_r = (1+\xi r)^{-1/2}$. This bounded mapping, which keeps every layer's effective rate in $(0, 2\eta_0\gamma_r)$ and re-centers persistent divergence back toward the global schedule, is what carries the argument: it converts locally observable optimization dynamics into per-layer step sizes without any extra messages.

What would settle it

Run FedA2L with the paper's default hyperparameters on a held-out architecture/dataset pair (for instance, a small transformer on CIFAR-100 with $\alpha=0.1$) and compare rounds-to-target against the best scheduler baseline; if FedA2L fails to converge or loses the speedup, the transferability premise is refuted. A sharper version sweeps $R_{\mathrm{warm}}$ over $\{10,40\}$ on that pair to see whether the bound reported in Table 6 (175 to 462 rounds on ResNet-18) reproduces on a second architecture.

Watch

Extended reading notes

Core claim

FedA2L's central claim is that the divergence between the three model states of a DFL round — base, trained, and aggregated — is enough information to assign useful per-layer learning rates. Weight divergence $\sigma_{i,l}^r$ measures how strongly local training moved layer $l$ at node $i$, and aggregation stability $\zeta_{i,l}^r$ measures how many of the layer's parameters survived neighbor aggregation nearly unchanged. After Z-score normalization over a sliding window, these two signals are fused into a bounded multiplier $1+\tanh(\log \lambda_{i,l}^r)$ that scales the base learning rate for that layer, with a global decay factor keeping rates on a diminishing schedule. The paper reports that this mechanism reaches target accuracy in fewer rounds than vanilla, scheduler-based, and static-search baselines in the large majority of configurations tested, with the biggest gains on deep architectures, severe non-IID splits, and sparse topologies where many baselines fail to converge. The theory section proves that bounded, slowly varying layer-wise multipliers do not change the $\mathcal{O}(1/\sqrt{R})$ non-convex rate or the linear rate in the strongly convex case; they only affect constants.

Load-bearing premise

The load-bearing premise is that the default FedA2L hyperparameters, selected on the CIFAR-100 benchmark, transfer to unseen architectures and workloads — yet Section 5.6 shows convergence can degrade sharply when $R_{\mathrm{warm}}$ or $\xi$ deviates from those defaults.

Editorial extensions

If this is right

  • FedA2L should reduce communication rounds to target accuracy by up to 59% relative to the best scheduler-based baseline, with the largest gains on deep networks and on sparse ring topologies where many baselines fail to converge.
  • Because FedA2L only modifies local step sizes, it can be dropped into existing DFL algorithms (FedAvg, FedProx, FedYogi, FedNTD, FedAWA, DFedSAM) without touching aggregation logic or adding communication overhead.
  • The bounded multiplier and sliding-window normalization guarantee that layer-wise learning rates stay within $(0, 2\eta_0\gamma_r)$ and converge back to the global decay schedule, so the method cannot amplify step sizes without bound.
  • The convergence analysis shows that FedA2L preserves the standard $\mathcal{O}(1/\sqrt{R})$ non-convex rate and the linear rate in the strongly convex case of the underlying DFL algorithm.
  • Under severe heterogeneity (Dirichlet $\alpha=0.01$), 20–30 nodes, and sparse topologies, FedA2L is reported to maintain or widen its advantage, including in cases where baselines do not converge at all.

Reading between the lines

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

  • Extension: the same three-state divergence signals could also drive layer-wise aggregation weights or per-layer communication compression, since they already summarize local drift and consensus mismatch without sending extra messages.
  • Extension: because Section 5.6 singles out $R_{\mathrm{warm}}$ and $\xi$ as the sensitive knobs, a short calibration phase that estimates them per architecture could make FedA2L nearly hyperparameter-free; the paper leaves this tuning to the user.
  • Extension: the reported gains shrink when the base algorithm already corrects drift (FedNTD, FedYogi), so a sensible deployment rule would engage FedA2L selectively, for base algorithms that lack such correction, rather than unconditionally.
  • Extension: an energy-accounting experiment that measures joules per round, rather than rounds to target, would directly test the edge-deployment motivation, since skipped rounds save both communication and battery.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes FedA2L, a server-free, layer-wise learning-rate adaptation method for decentralized federated learning. At each node, FedA2L derives two per-layer metrics from the three intra-round model states: weight divergence (local update intensity) and aggregation stability (network consensus), normalizes them with Z-scores over a sliding window, fuses them into a score, and modulates the base learning rate with a bounded factor and a global polynomial decay. The method adds no communication overhead and is integrated into existing DFL protocols. The paper reports extensive experiments across six DFL algorithms, three image architectures, five datasets, wall-clock convergence, robustness to heterogeneity, scalability, topologies, and time-series forecasting, together with a non-convex convergence theorem. The main claimed results are up to 4.94x faster convergence than vanilla DFL and up to 59% fewer communication rounds than the best scheduler-based baselines. The authors also release code.

Significance. If the central claim holds, FedA2L is a practically attractive contribution: it is local-only, communication-free, easy to plug into existing DFL pipelines, and the empirical campaign is unusually broad for a DFL learning-rate method. The breadth of the evaluation (six base algorithms, three architectures, five datasets, wall-clock and round-based metrics, multiple topologies) is a genuine strength, as is the explicit release of code. The convergence result, if fully established, would show that the modification does not degrade the standard decentralized SGD rate. However, the significance is currently conditional on three issues: the evidence does not isolate layer-wise adaptation from a global learning-rate scale effect; several key hyperparameters are tuned on the same CIFAR-100 benchmark used for headline evaluation; and the convergence analysis rests on an unproved slow-variation assumption plus sketched proof steps. These are load-bearing for the paper's central claims, so the contribution is promising but not yet fully supported.

major comments (4)
  1. [§3.4.3, Eqs. (7)-(8); §5.6] The paper does not isolate the claimed layer-wise mechanism from a global learning-rate scale effect. The multiplier m_{i,l}^r = 1+tanh(log(beta e^{omega} + (1-beta)e^{delta})) is not mean-zero over layers: by weighted AM-GM, log(beta e^a + (1-beta)e^b) >= beta a + (1-beta)b, and the Z-scores are centered, so the fusion score is systematically right-skewed and the layer-averaged multiplier can exceed 1, especially in early rounds. The paper never reports the mean multiplier trajectory and never compares against a control that applies a single common multiplier to all layers while preserving the average scale. The ablations in Table 6 and Fig. 5 vary beta and other hyperparameters, but both arms retain layer-wise variation, so they cannot distinguish per-layer allocation from a common scale boost. Please add a control experiment with a single global multiplier mbar^r = (1/L) sum_l m_{i,l}^r applied uniformly, and report the trajectory of the mean multiplier; if the speedup persists in that control, the central 'layer-wise' mechanism claim needs to be substantially revised.
  2. [§5.6.2, Table 7; §5.1] The default hyperparameters are selected on CIFAR-100 and then used in the main evaluation that includes CIFAR-100, so the headline numbers are not an independent test of the method's robustness. Table 6 shows that the defaults are consequential: R_warm=10 on ResNet-18 raises rounds to target from 175 to 462, and xi=0.3 causes non-convergence. Table 7 then fixes R_warm and xi per architecture based on that sensitivity analysis, and even adds a post-hoc adjustment (xi=0.05 for TinyImageNet) after seeing that dataset. This is tuning on the evaluation benchmark. Please provide a holdout evaluation protocol: select defaults on one benchmark (e.g., CIFAR-100), then evaluate on the remaining datasets with those defaults fixed, or use nested validation, and report the sensitivity of the headline claims to that choice.
  3. [§4, Assumption A5 and Theorem 1] The convergence analysis relies on an unproved slow-variation property. Assumption A5 requires |m^{r+1}-m^r| <= Delta_m, and the text asserts that Z-score normalization over a finite window, together with bounded tanh, 'ensures' this. Boundedness of m in (0,2) does not by itself bound round-to-round changes: Z-scores can jump when the sliding-window standard deviation is near epsilon, and the paper provides no lemma controlling Delta_m. Since Delta_m enters Eq. (15) through the term C3 L eta_max^2 Delta_m^2/phi^2, the claimed O(1/sqrt R) rate depends on Delta_m being a constant independent of R. In addition, the proof of Theorem 1 is only a sketch: the disagreement recursion for D^r with layer-wise heterogeneous step sizes is asserted, not derived, and the strongly convex result in §4.2 is stated without proof. The theoretical section needs a formal lemma establishing A5, or an alternative argument that does not depend on unproved slow variation, plus full proof details for the stated theorems.
  4. [§5.1.3 with Tables 4, 5, 6] The statistical reporting is inconsistent with the stated protocol. Section 5.1.3 says all results are mean values across five independent runs, but Tables 4, 5, and 6 report single numbers without standard deviations or confidence intervals. Table 2 itself shows large baseline variability (e.g., FedProx CAWR on CNN/CIFAR-10: 148 ± 64.78 rounds), so unquantified comparisons in the robustness, scalability, topology, and time-series tables do not support the corresponding claims at the stated level of rigor. Please report mean ± std (or CIs) for all tables, at least for the FedA2L rows and the best baseline in each setting.
minor comments (5)
  1. [Algorithm 1] The pseudocode does not explicitly initialize the learning-rate vector before the warm-up rounds; please state that eta_i^r = eta0 for all layers during r <= R_warm.
  2. [§3.4.3] The phrase 'When the fusion score lambda=1 (typical behavior)' is presented as a fact, but the paper does not demonstrate that lambda=1 is typical; given the skew of the fusion score discussed in the major comments, this statement should be either justified empirically or rephrased.
  3. [Table 4] The table header lists target accuracies (40%, 45%, 42%, 40%, 33%, 45%) but the column groups are not labeled in the displayed text; please add explicit column headers for Heterogeneity, Scalability, and Topology to improve readability.
  4. [§5.6.1, Fig. 4] The right-axis curves are described as 'linearly rescaled to [0,1] for visualization,' but the raw scales of the metrics are not reported; please provide the actual ranges or a representative example so the reader can assess the magnitude of the divergence signals.
  5. [§4.1, Eq. (15)] The constants C1, C2, C3 are stated to be independent of R but their dependencies on L, phi, and the multiplier bounds are not specified; please state them explicitly or give their definitions in the proof.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Theorem 1 relies on assumptions satisfied by the method's own bounded-tanh construction, and the headline speedups are empirical benchmark measurements, not quantities forced by fitted inputs.

full rationale

FedA2L's central claims are empirical: communication rounds to target accuracy are measured against vanilla and scheduler baselines under a fixed protocol, so the reported 4.94x and 59% figures are benchmark observations, not values derivable from the method's equations by construction. The convergence analysis in Section 4 does not smuggle in the conclusion: Assumption A5 requires bounded and slowly varying multipliers, and the paper's own construction guarantees this; lambda>0 in Eq. 7 implies m=1+tanh(log(lambda)) lies in (0,2) from Eq. 8, so both boundedness and a finite Delta_m follow immediately without needing the Z-score window. Theorem 1 then only asserts preservation of the standard O(1/sqrt(R)) rate, a consistency statement rather than an explanation of the observed speedups. Self-citations (refs 15, 33, 38) appear only in background/context and are not load-bearing; no uniqueness theorem or ansatz is imported from the authors' prior work. The hyperparameter defaults in Table 7 were selected on CIFAR-100 and then used in the CIFAR-100 comparisons, and xi was adjusted for TinyImageNet; this is a potential tuning/leakage concern that could bias the empirical comparison, but it is not circularity because the reported convergence rounds are not algebraically forced by those parameter choices. Two non-circular weaknesses are worth noting: the narrative in Section 3.4.3 states that weak consensus triggers a higher LR, while Eqs. 7-8 use zeta directly, so a high zeta raises e^{delta} and therefore the multiplier, which is an internal sign inconsistency; and the ablations vary beta without a global-mean-multiplier control, so the layer-wise attribution of the speedup is not isolated from a possible global-LR-scale effect. These are correctness and experimental-design concerns, not cases where a prediction reduces to its input by definition.

Assumptions & free parameters 6 free parameters · 7 assumptions · 0 invented entities

The central empirical claim depends on hyperparameters tuned on the headline benchmark and on the layer-divergence heuristic. The convergence theorem depends on standard decentralized optimization assumptions plus A5, which is asserted rather than proven. No new physical or algorithmic entities are postulated.

free parameters (6)
  • R_warm (warm-up rounds) = 10 (CNN), 40 (ResNet)
    Chosen from {10,20,40} in Section 5.6; R_warm=10 degrades ResNet-18 rounds from 175 to 462.
  • rho (Z-score window) = 5 (CNN), 10 (ResNet)
    Chosen from {5,10,20}; sensitivity is modest but the reported default is tuned per architecture.
  • tau (stability threshold) = 1e-3
    Tuned from {1e-2,1e-3,1e-4}; tau=1e-4 degrades ResNet convergence from 171 to 192 rounds.
  • xi (global decay constant) = 0.3 (CNN), 0.1 (ResNet), 0.05 (TinyImageNet)
    Selected from {0.05,0.1,0.3}; xi=0.3 causes convergence failure on ResNet-18 and ResNet-34.
  • beta (fusion balance) = 0.6
    Selected from {0.0,0.2,0.4,0.6,0.8,1.0}; stable in [0.4,0.8], extremes beta=0 and beta=1 ablate the two signals.
  • eta0 (base learning rate) = 0.01
    Initial LR used across all experiments; chosen by hand, not derived.
assumptions (7)
  • standard math A1: each local loss is L-smooth
    Standard smoothness assumption used in Section 4 for Theorem 1.
  • standard math A2: stochastic gradients are unbiased with bounded variance
    Standard SGD assumption in Section 4, Assumption A2.
  • domain assumption A3: mixing matrices are doubly stochastic with a spectral gap
    Standard decentralized optimization assumption in Section 4, Assumption A3.
  • domain assumption A4: gradient heterogeneity is bounded by B
    Standard non-IID bound in Section 4, Assumption A4.
  • ad hoc to paper A5: multipliers m are bounded and slowly varying
    Assumed for the theorem; the paper asserts Z-score normalization implies |m^{r+1}-m^r| <= Delta_m, but gives no proof. This is the key condition that keeps the method in the standard convergence rate class.
  • domain assumption Layer hierarchy premise: foundational vs specialized layers require different learning rates
    Section 3.2 motivates the entire method through this claim; it is argued intuitively, not derived.
  • domain assumption Fixed stability threshold tau is meaningful across layers
    Equation 4 uses an absolute threshold on parameter changes without normalization by layer parameter scale, so transfer across architectures is assumed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FedA2L: Adaptive layer-wise learning rate adjustment in decentralized federated learning." pith.science (2026). https://pith.science/paper/TEFTEPFW

@misc{pith2026260809208,
  author       = {Pith},
  title        = {Pith review of: FedA2L: Adaptive layer-wise learning rate adjustment in decentralized federated learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TEFTEPFW}},
  note         = {Machine review of arXiv:2608.09208}
}
read the original abstract

Decentralized intelligence systems with heterogeneous devices and limited coordination increasingly rely on decentralized federated learning (DFL). However, DFL suffers from convergence inefficiency under data heterogeneity due to the use of a uniform learning rate (LR) that ignores layer-specific optimization needs. Foundational layers are responsible for maintaining network consensus, while specialized layers adapt to local data characteristics, leading to conflicting gradients and degraded performance under non-IID conditions. To address this fundamental tension, this work introduces FedA2L, a method that dynamically adjusts layer-wise LRs based on model divergence signals. By leveraging local update intensity and network consensus constraints, FedA2L seamlessly integrates into existing DFL protocols without additional communication or coordination. Extensive evaluations across DFL algorithms, various model architectures, and datasets demonstrate that FedA2L achieves up to 4.94 times faster convergence than vanilla DFL and reduces communication rounds by up to 59% compared to scheduler-based baselines. Furthermore, FedA2L exhibits resilience to severe data heterogeneity, larger network sizes, and sparse topologies, reducing communication overhead and establishing it as a versatile optimization tool for resource-constrained or large-scale distributed learning in edge and IoT deployments. The code is released at https://github.com/nclabteam/FedA2L.

Figures

Figures reproduced from arXiv: 2608.09208 by the authors.

Figure 1
Figure 1. (a) Standard DFL framework with three intra-round model states per layer (𝜃 𝐵,𝑟 𝑖,𝑙 , 𝜃 𝑇 ,𝑟 𝑖,𝑙 , 𝜃 𝐴,𝑟 𝑖,𝑙 ). (b) FedA2L integrated into the DFL framework. During each round, the three states of layer 𝑙 are used to compute two layer-wise metrics, 𝜎 𝑟 𝑖,𝑙 (Eq. 3) and 𝜁 𝑟 𝑖,𝑙 (Eq. 4), which in turn determine the adaptive layer-wise learning rate vector 𝜂 𝑟+1 𝑖 = [𝜂𝑖,1 , …, 𝜂𝑖,𝐿] 𝑟+1 for the next round. data, we exam… view at source ↗
Figure 2
Figure 2. Time to reach 45% target accuracy on CIFAR-100 using ResNet-18 (Dirichlet 𝛼 = 0.1) across six DFL algorithms. while narrowing scheduler gaps rather than delivering the strongest speedups in every configuration. These results indi￾cate that FedA2L’s benefits increase with feature hierarchy depth and representation complexity, rather than with network depth alone. 5.2.5. Comparison of convergence time FedA2L consisten… view at source ↗
Figure 3
Figure 3. Highest test accuracy achieved across algorithms, models, and datasets under non-IID partitioning (Dirichlet 𝛼 = 0.1) [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: LR feedback dynamics for the three layers with the highest local divergence (ResNet-18, CIFAR-100, FedAvg, Dirichlet 𝛼 = 0.1). The plots show the layer-wise learning rate 𝜂 𝑟 𝑖,𝑙 (left axis), together with the normalized weight divergence 𝜎 𝑟 𝑖,𝑙 and aggregation instab…
Figure 5
Figure 5. Figure 5: Effect of fusion balance 𝛽 on convergence speed and model performance under FedAvg on CIFAR-100. The extreme values 𝛽 = 0 (𝜁-only) and 𝛽 = 1 (𝜎-only) constitute direct component ablations of 𝜁 and 𝜎, respectively, in Eq. 7, while intermediate values reflect the full du…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

56 extracted references · 40 canonical work pages

  1. [1]

    Adaptive federated learning and digital twin for industrial internet of things

    WenSun,ShiyuLei,LuWang,ZhiqiangLiu,andYanZhang. Adaptive federated learning and digital twin for industrial internet of things. IEEE Transactions on Industrial Informatics, 17(8):5605–5614, 2021. doi: 10.1109/TII.2020.3034674

    arXiv 2021

  2. [2]

    Digital twin driven smart factories: real time physics based co-simulation using edge ai and federated learning.Scientific Reports, 15(1):43373, 2025

    V Padmavathi, R Kanimozhi, and R Saminathan. Digital twin driven smart factories: real time physics based co-simulation using edge ai and federated learning.Scientific Reports, 15(1):43373, 2025

    work page 2025

  3. [3]

    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. InArtificial intelligence and statistics, pages 1273–1282. PMLR, 2017

    2017

  4. [4]

    Edge-basedcommunicationoptimizationfordistributedfederated learning.IEEE Transactions on Network Science and Engineering, 9 (4):2015–2024, 2021

    TianWang,YanLiu,XiZheng,Hong-NingDai,WeijiaJia,andMande Xie. Edge-basedcommunicationoptimizationfordistributedfederated learning.IEEE Transactions on Network Science and Engineering, 9 (4):2015–2024, 2021

    work page 2015

  5. [5]

    ShuaiWang,YouliangTian,JinboXiong,JianfengMa,andYanZhang. VerifyDFL: Secure aggregation for decentralized federated learning with input validation in mobile edge intelligence.IEEE Transactions on Cognitive Communications and Networking, 12:2526–2541, 2026. doi: 10.1109/TCCN.2025.3587771

    work page arXiv 2026

  6. [6]

    Yu, and Christo- pher G

    Liangqi Yuan, Ziran Wang, Lichao Sun, Philip S. Yu, and Christo- pher G. Brinton. Decentralized federated learning: A survey and perspective.IEEE Internet of Things Journal, 11(21):34617–34638,

    work page

  7. [7]

    Fullydecentralizedfederatedlearning

    Anusha Lalitha, Shubhanshu Shekhar, Tara Javidi, and Farinaz Koushanfar. Fullydecentralizedfederatedlearning. InThirdworkshop on bayesian deep learning (NeurIPS), volume 12, 2018

    work page 2018

  8. [8]

    Distributedandsecurefederatedlearningforwirelesscomputingpower networks.IEEE Transactions on Vehicular Technology, 72(7):9381– 9393, 2023

    Peng Wang, Wen Sun, Haibin Zhang, Wenqiang Ma, and Yan Zhang. Distributedandsecurefederatedlearningforwirelesscomputingpower networks.IEEE Transactions on Vehicular Technology, 72(7):9381– 9393, 2023

    work page 2023

Show all 56 references
  1. [9]

    MC-2PF:Amulti-edgecooperativeuniversalframework for load prediction with personalized federated deep learning.IEEE Transactions on Mobile Computing, 24(6):5138–5154, 2025

    Zheyi Chen, Qingnan Jiang, Lixian Chen, Xing Chen, Jie Li, and GeyongMin. MC-2PF:Amulti-edgecooperativeuniversalframework for load prediction with personalized federated deep learning.IEEE Transactions on Mobile Computing, 24(6):5138–5154, 2025. doi: 10.1109/TMC.2025.3528404

    2025

  2. [10]

    Resilientcollaborativecachingformulti-edgesystemswith robust federated deep learning.IEEE Transactions on Networking, 33 (2):654–669, 2024

    Zheyi Chen, Jie Liang, Zhengxin Yu, Hongju Cheng, Geyong Min, andJieLi. Resilientcollaborativecachingformulti-edgesystemswith robust federated deep learning.IEEE Transactions on Networking, 33 (2):654–669, 2024

    2024

  3. [11]

    Accelerating decentralized federated learning with probabilistic communication in heterogeneous edgecomputing.IEEETransactionsonNetworking,34:486–501,2026

    Jianchun Liu, Jiaming Yan, Hongli Xu, Lun Wang, Zhiyuan Wang, Jinyang Huang, and Chunming Qiao. Accelerating decentralized federated learning with probabilistic communication in heterogeneous edgecomputing.IEEETransactionsonNetworking,34:486–501,2026. doi: 10.1109/TON.2025.3600015

    2026

  4. [12]

    Decentralized edge learning: A comparative study of distillation strategies and dissimilarity measures.Future Generation Computer Systems, 176:108171, 2026

    Mbasa Joaquim Molo, Lucia Vadicamo, Claudio Gennaro, and Emanuele Carlini. Decentralized edge learning: A comparative study of distillation strategies and dissimilarity measures.Future Generation Computer Systems, 176:108171, 2026. ISSN 0167-739X. doi: https://doi.org/10.1016/...

    2026

  5. [13]

    Traffic-Aware Lightweight Hierarchical Offloading Toward Adaptive Slicing-EnabledSAGIN.IEEEJournalonSelectedAreasinCommuni- cations, 42(12):3536–3550, 2024

    Zheyi Chen, Junjie Zhang, Geyong Min, Zhaolong Ning, and Jie Li. Traffic-Aware Lightweight Hierarchical Offloading Toward Adaptive Slicing-EnabledSAGIN.IEEEJournalonSelectedAreasinCommuni- cations, 42(12):3536–3550, 2024. doi: 10.1109/JSAC.2024.3459020

    2024

  6. [14]

    Federated optimization in heterogeneous networks.Proceedings of Machine learning and systems, 2:429–450, 2020

    TianLi,AnitKumarSahu,ManzilZaheer,MaziarSanjabi,AmeetTal- walkar, and Virginia Smith. Federated optimization in heterogeneous networks.Proceedings of Machine learning and systems, 2:429–450, 2020

    2020

  7. [15]

    FedTVD:Balancing data quality and quantity for robust federated learning.Future Generation Computer Systems, page 108177, 2025

    RadwanSelo,MajidKundroo,andTaehongKim. FedTVD:Balancing data quality and quantity for robust federated learning.Future Generation Computer Systems, page 108177, 2025

    2025

  8. [16]

    FedDC: Federated learning with non-IID data via local drift decoupling and correction

    Liang Gao, Huazhu Fu, Li Li, Yingwen Chen, Ming Xu, and Cheng- Zhong Xu. FedDC: Federated learning with non-IID data via local drift decoupling and correction. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10112– 10121, 2022

    2022

  9. [17]

    Def-Ag: An energy- efficient decentralized federated learning framework via aggregator clients.Future Generation Computer Systems, 175:108114, 2026

    Junyoung Park, Sungpil Woo, and Joohyung Lee. Def-Ag: An energy- efficient decentralized federated learning framework via aggregator clients.Future Generation Computer Systems, 175:108114, 2026. ISSN 0167-739X. doi: https://doi.org/10.1016/j.future.2025.108114

    2026

  10. [18]

    How transferable are features in deep neural networks?Advances in neural information processing systems, 27, 2014

    Jason Yosinski, Jeff Clune, Yoshua Bengio, and Hod Lipson. How transferable are features in deep neural networks?Advances in neural information processing systems, 27, 2014

    2014

  11. [19]

    Visualizing and understanding convolutional networks

    Matthew D Zeiler and Rob Fergus. Visualizing and understanding convolutional networks. InComputer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part I 13, pages 818–833. Springer, 2014

    2014

  12. [20]

    PLayer-FL: A principled approachtopersonalizedlayer-wisecross-silofederatedlearning,2025

    Ahmed Elhussein and Gamze Gürsoy. PLayer-FL: A principled approachtopersonalizedlayer-wisecross-silofederatedlearning,2025. URLhttps://arxiv.org/abs/2502.08829

    2025 arXiv

  13. [21]

    Optimizing personalized federated learning through adaptive layer- wise learning

    Weihang Chen, Cheng Yang, Jie Ren, Zhiqiang Li, and Zheng Wang. Optimizing personalized federated learning through adaptive layer- wise learning. In James Kwok, editor,Proceedings of the Thirty- FourthInternationalJointConferenceonArtificialIntelligence,IJCAI- 25, pages 4860–4...

    2025

  14. [22]

    Adaptive federated optimization.arXiv preprint arXiv:2003.00295, 2021

    Sashank Reddi, Zachary Charles, Manzil Zaheer, Zachary Garrett, KeithRush,JakubKonečn `y,SanjivKumar,andHBrendanMcMahan. Adaptive federated optimization.arXiv preprint arXiv:2003.00295, 2021

    2003 arXiv

  15. [23]

    FLARE: A new federated learning framework with adjustable learning rates over resource-constrained wireless networks.IEEE Transactions on Wireless Communications, 2025

    Bingnan Xiao, Jingjing Zhang, Wei Ni, and Xin Wang. FLARE: A new federated learning framework with adjustable learning rates over resource-constrained wireless networks.IEEE Transactions on Wireless Communications, 2025

    2025

  16. [24]

    Consensus-driven hyperparameter optimization for accelerated model convergence in decentralized federated learning

    Anam Nawaz Khan, Qazi Waqas Khan, Atif Rizwan, Rashid Ahmad, and Do Hyeun Kim. Consensus-driven hyperparameter optimization for accelerated model convergence in decentralized federated learning. Internet of Things, 30:101476, 2025. ISSN 2542-6605. doi: https: //doi.org/10.1016...

    2025

  17. [25]

    AutoLR: Layer-wise pruning and auto-tuning of learning rates in fine-tuning of deep networks

    Youngmin Ro and Jin Young Choi. AutoLR: Layer-wise pruning and auto-tuning of learning rates in fine-tuning of deep networks. InProceedings of the AAAI Conference on Artificial Intelligence, volume 35, pages 2486–2494, 2021

    2021

  18. [26]

    Fed-LAMB: layer-wise and dimension-wise locally adaptive federated learning

    Belhal Karimi, Ping Li, and Xiaoyun Li. Fed-LAMB: layer-wise and dimension-wise locally adaptive federated learning. InUncertainty in Artificial Intelligence, pages 1037–1046. PMLR, 2023

    2023

  19. [27]

    FedLWS:Federatedlearningwithadaptivelayer-wiseweightshrinking

    Changlong Shi, Jinmeng Li, He Zhao, Dandan Guo, and Yi Chang. FedLWS:Federatedlearningwithadaptivelayer-wiseweightshrinking. arXiv preprint arXiv:2503.15111, 2025

    2025 arXiv

  20. [28]

    Xiangru Lian, Ce Zhang, Huan Zhang, Cho-Jui Hsieh, Wei Zhang, and Ji Liu. Can decentralized algorithms outperform centralized algorithms? a case study for decentralized parallel stochastic gradient descent.Advancesinneuralinformationprocessing systems,30,2017

    2017

  21. [29]

    Decentralized federated learning: Fundamentals, state of the art, frameworks, trends, and challenges.IEEE Communications Surveys & Tutorials, 25(4):2983–3013, 2023

    Enrique Tomás Martínez Beltrán, Mario Quiles Pérez, Pedro Miguel Sánchez Sánchez, Sergio López Bernal, Gérôme Bovet, Manuel Gil Pérez, Gregorio Martínez Pérez, and Alberto Huertas Celdrán. Decentralized federated learning: Fundamentals, state of the art, frameworks, trends, an...

    2023

  22. [30]

    Enhancing privacy preservation and trustworthiness for decentralized federated learning.Information Sciences, 628:449–468,

    Lingling Wang, Xueqin Zhao, Zhongkai Lu, Lin Wang, and Shouxun Zhang. Enhancing privacy preservation and trustworthiness for decentralized federated learning.Information Sciences, 628:449–468,

  23. [31]

    DeFTA:Aplug-and-playpeer-to-peerdecentralizedfederatedlearning framework.InformationSciences,670:120582,2024.ISSN0020-0255

    Yuhao Zhou, Minjia Shi, Yuxin Tian, Qing Ye, and Jiancheng Lv. DeFTA:Aplug-and-playpeer-to-peerdecentralizedfederatedlearning framework.InformationSciences,670:120582,2024.ISSN0020-0255. doi: https://doi.org/10.1016/j.ins.2024.120582

    2024

  24. [32]

    Network topology and communication-computation tradeoffs in decentralized optimization.Proceedings of the IEEE, 106(5):953–976, 2018

    Angelia Nedić, Alex Olshevsky, and Michael G Rabbat. Network topology and communication-computation tradeoffs in decentralized optimization.Proceedings of the IEEE, 106(5):953–976, 2018

    2018

  25. [33]

    Vo van Truong, Pham Khanh Quan, Dong-Hwan Park, and Taehong Kim. Performance evaluation of decentralized federated learning: Impact of fully and k-connected topologies, heterogeneous computing resources, and communication bandwidth.IEEE Access, 13:32741– 32755, 2025. doi: 10.1...

    2025

  26. [34]

    Stich, and Martin Jaggi

    Anastasia Koloskova, Sebastian U. Stich, and Martin Jaggi. Decentral- ized stochastic optimization and gossip algorithms with compressed communication. InProceedings of the 36th International Conference on Machine Learning (ICML), volume 97 ofProceedings of Machine Learning Re...

    2019

  27. [35]

    GossipFL:A decentralized federated learning framework with sparsified and adap- tive communication.IEEE Transactions on Parallel and Distributed Systems, 34(3):909–922, 2022

    ZhenhengTang,ShaohuaiShi,BoLi,andXiaowenChu. GossipFL:A decentralized federated learning framework with sparsified and adap- tive communication.IEEE Transactions on Parallel and Distributed Systems, 34(3):909–922, 2022

    2022

  28. [36]

    FedAWA: Adaptive optimization of aggregation weights in federated learning using client vectors

    Changlong Shi, He Zhao, Bingjie Zhang, Mingyuan Zhou, Dandan Guo, and Yi Chang. FedAWA: Adaptive optimization of aggregation weights in federated learning using client vectors. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 30651–30660, 2025

    2025

  29. [37]

    Adapted weighted aggregation in federated learning

    Yitong Tang. Adapted weighted aggregation in federated learning. InProceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 23763–23765, 2024

    2024

  30. [38]

    Federated learning with hyper- parameter optimization.Journal of King Saud University-Computer and Information Sciences, 35(9):101740, 2023

    Majid Kundroo and Taehong Kim. Federated learning with hyper- parameter optimization.Journal of King Saud University-Computer and Information Sciences, 35(9):101740, 2023

    2023

  31. [39]

    Federated hyperparameter optimization through reward-basedstrategies:Challengesandinsights

    Krishna Kanth Nakka, Ahmed Frikha, Ricardo Mendis, Xue Jiang, and Xuebing Zhou. Federated hyperparameter optimization through reward-basedstrategies:Challengesandinsights. InProceedingsofthe IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4236–4244, 2024

    2024

  32. [40]

    Federated learning with non-IID data.arXiv preprint arXiv:1806.00582, 2018

    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

    2018 arXiv

  33. [41]

    Federatedlearningvia consensus mechanism on heterogeneous data: A new perspective on convergence

    ShuZheng,TiandiYe,XiangLi,andMingGao. Federatedlearningvia consensus mechanism on heterogeneous data: A new perspective on convergence. InICASSP2024-2024IEEEInternationalConferenceon Acoustics, Speech and Signal Processing (ICASSP), pages 7595–7599. IEEE, 2024

    2024

  34. [42]

    Learning multiple layers of features from tiny images

    Alex Krizhevsky and Geoffrey Hinton. Learning multiple layers of features from tiny images. Technical Report 0, University of Toronto, Toronto, Ontario, 2009

    2009

  35. [43]

    Tiny imagenet visual recognition challenge

    Yann Le and Xuan Yang. Tiny imagenet visual recognition challenge. CS 231N, 7(7):3, 2015

    2015

  36. [44]

    Guokun Lai, Wei-Cheng Chang, Yiming Yang, and Hanxiao Liu. Modeling long-and short-term temporal patterns with deep neural networks.InThe41stinternationalACMSIGIRconferenceonresearch & development in information retrieval, pages 95–104, 2018

    2018

  37. [45]

    Cautionary tales on air-quality improvement in beijing

    Shuyi Zhang, Bin Guo, Anlan Dong, Jing He, Ziping Xu, and Song Xi Chen. Cautionary tales on air-quality improvement in beijing. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 473(2205), 2017

    2017

  38. [46]

    Timer: generative pre-trained trans- formers are large time series models

    Yong Liu, Haoran Zhang, Chenyu Li, Xiangdong Huang, Jianmin Wang, and Mingsheng Long. Timer: generative pre-trained trans- formers are large time series models. InProceedings of the 41st International Conference on Machine Learning, ICML’24. JMLR.org, 2024

    2024

  39. [47]

    Preser- vation of the global knowledge by not-true self knowledge distillation in federated learning.CoRR, abs/2106.03097, 2021

    Gihun Lee, Yongjin Shin, Minchan Jeong, and Se-Young Yun. Preser- vation of the global knowledge by not-true self knowledge distillation in federated learning.CoRR, abs/2106.03097, 2021

    2021 arXiv

  40. [48]

    Improving the model consistency of decentralized federated learning

    Yifan Shi, Li Shen, Kang Wei, Yan Sun, Bo Yuan, Xueqian Wang, and Dacheng Tao. Improving the model consistency of decentralized federated learning. InProceedings of the 40th International Confer- ence on Machine Learning, volume 202 ofProceedings of Machine Learning Research, ...

    2023

  41. [49]

    SGDR: Stochastic gradient descent with warm restarts, 2017

    Ilya Loshchilov and Frank Hutter. SGDR: Stochastic gradient descent with warm restarts, 2017. URLhttps://arxiv.org/abs/1608.03983

    2017 arXiv

  42. [50]

    Smith and Nicholay Topin

    Leslie N. Smith and Nicholay Topin. Super-convergence: Very fast training of neural networks using large learning rates, 2018. URL https://arxiv.org/abs/1708.07120

    2018 arXiv

  43. [51]

    HyperbolicLR: Epoch insensitive learning rate scheduler, 2025

    Tae-Geun Kim. HyperbolicLR: Epoch insensitive learning rate scheduler, 2025. URLhttps://arxiv.org/abs/2407.15200

    2025 arXiv

  44. [52]

    FedMMD: A Federated weighting algorithm consideringNon-IIDandLocalModelDeviation.ExpertSystemswith Applications, 237:121463, 2024

    KaiHu,YaogenLi,ShuaiZhang,JiashengWu,ShengGong,Shanshan Jiang, and Liguo Weng. FedMMD: A Federated weighting algorithm consideringNon-IIDandLocalModelDeviation.ExpertSystemswith Applications, 237:121463, 2024

    2024

  45. [53]

    Personalized federated learning via deviation tracking representation learning

    Jaewon Jang and Bong Jun Choi. Personalized federated learning via deviation tracking representation learning. In2024 International ConferenceonInformationNetworking(ICOIN),pages762–766.IEEE, 2024

    2024

  46. [54]

    FedRDA: Representation Deviation Alignment in Heterogeneous Federated Learning.IEEE Transactions on Industrial Informatics, 2025

    Wenjie Yao, Guanglu Sun, Suxia Zhu, Ruidong Wang, Xinzhong Zhu, HuiYing Xu, and Xiguang Wei. FedRDA: Representation Deviation Alignment in Heterogeneous Federated Learning.IEEE Transactions on Industrial Informatics, 2025. Truong et al.:Preprint submitted to ElsevierPage 18 of 18

    2025

  47. [2023]

    doi:https://doi.org/10.1016/j.ins.2023.01.130

    ISSN0020-0255. doi:https://doi.org/10.1016/j.ins.2023.01.130

    2023 doi

  48. [2024]

    doi: 10.1109/JIOT.2024.3407584

    2024

Pith tools

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