Pith. sign in

REVIEW 3 major objections 5 minor 48 references

An Information-Theoretic Analysis for Federated Learning under Concept Drift

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper claims that the long-term performance of federated learning under concept drift is governed by an information-theoretic bound on a Stationary Generalization Error, and that this bound correctly orders the damage done by random…

desk verdict Genuinely new SGE metric and pattern bounds, but Theorem 1 is unproven for the paper's own losses because the proof silently assumes Bregman divergence. read the letter →

arxiv 2506.21036 v1 pith:KYOGJCWF submitted 2025-06-26 cs.LG cs.DC

Fu Peng , Meng Zhang , Ming Tang This is my paper · ORCID
classification cs.LGcs.DC MSC 68T0594A17
keywords federatedlearningconceptdriftStationaryGeneralizationErrorMarkovchainmutualinformationKLdivergenceperformance-costtradeoffinformation-theoreticbound
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 an information-theoretic account of how concept drift degrades federated learning, and to turn that account into a usable training rule. It models the evolving data distribution as a Markov chain and defines the Stationary Generalization Error, the steady-state expected gap between population and empirical risk across a current and a future unseen distribution. Theorem 1 bounds this error by mutual information between the learned model and individual samples plus KL divergences between consecutive distributions; dropping the mutual information under a large-sample assumption yields closed-form bounds for periodic, gradual, and random drift (Corollary 1) and a threshold ordering in which random drift is worst, periodic intermediate, and gradual mildest when drift is drastic (Proposition 1). The paper then proposes an algorithm that regularizes empirical risk minimization with KL divergence on representation distributions and Gaussian noise, and reports that it outperforms four baselines on Fashion-MNIST and CIFAR-10 across all three patterns. The same bound is used to derive a closed-form, convex Pareto frontier between the Stationary Generalization Error and system cost.

What carries the argument

The load-bearing object is the Stationary Generalization Error $G$, defined as the steady-state average, over a Markov chain of data distributions, of a weighted generalization error that mixes the model's fit to the current distribution (stage I) and its fit to the next unseen distribution (stage II). The argument runs through the upper bound of Theorem 1: using the Legendre dual $\psi^{*-1}$ of the cumulant-generating function of the loss, each term of $G$ is bounded by $\psi^{*-1}$ applied to a mutual information term plus a KL divergence between distribution states. With sub-Gaussian losses, $\psi^{*-1}(x)=\sqrt{2r^2x}$, and with the large-sample assumption $I(w_{\mathrm{cur}},Z)\to 0$, the bound collapses to a weighted sum of square-root KL divergences driven only by the Markov chain's transition probabilities. The periodic, gradual, and random drift patterns are encoded as three state machines, and the closed-form bounds of Corollary 1 come from substituting their stationary distributions into that simplified bound. The algorithm then acts on the same two knobs: KL regularization on representation distributions lowers the divergence terms, and injected Gaussian noise lowers the mutual information term.

What would settle it

Recompute the bounds in Eqs. (15)-(17) without setting $I(w_{\mathrm{cur}},Z)$ to zero, using the $N=600$ sample size and the actual trained models from the testbed; if the mutual information terms are comparable to the KL terms, the predicted pattern ordering $B_r>B_p>B_g$ for small $p$ need not hold.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that the population risk of a federated model deployed under drifting data is not an unpredictable quantity but is controlled by two information-theoretic quantities: how much the retrained model $w_{\mathrm{cur}}$ depends on each individual training sample (mutual information $I(w_{\mathrm{cur}}, Z)$), and how far the previous, current, and next data distributions are from each other (KL divergences). Theorem 1 packages this in the Stationary Generalization Error $G$, averaged over the steady state of a Markov chain over distribution states. In the large-sample limit, the mutual information terms vanish and the bound becomes a sum of square-root KL distances weighted by transition probabilities; evaluating that sum for the three canonical drift state machines yields the explicit bounds in Corollary 1 and the threshold ordering in Proposition 1. The paper further claims that this analysis supports a practical training rule, Algorithm 1, that regularizes empirical risk minimization with symmetric KL matching of learned representations plus Gaussian noise, and that the same bound produces a closed-form convex Pareto frontier for the performance-cost tradeoff (Theorem 2).

Load-bearing premise

The ordering of drift patterns in the simplified bounds assumes the sample size is large enough that each single training sample's influence on the learned model, the mutual information term $I(w_{\mathrm{cur}},Z)$, is zero; the experiments use 600 samples per client, so Eqs. (13)-(17) may not hold exactly at that size.

Editorial extensions

If this is right

  • If Theorem 1 is correct, retraining algorithms can be compared by two quantities alone: how much the model depends on a single training sample ($I(w_{\mathrm{cur}},Z)$) and how far consecutive data distributions are in KL divergence.
  • The simplified bound predicts that for small transition probability $p$ (drastic drift), random drift produces the largest stationary generalization error, periodic drift is intermediate, and gradual drift is the mildest; for larger $p$ the periodic ordering can reverse.
  • Proposition 2 directs system design: placing more weight on fresh data (larger $\alpha$), using more clients (larger $K$), and shortening the adaptation window (smaller $\tau_2$) all shrink the upper bound.
  • Theorem 2 gives a closed-form, convex Pareto frontier, so an operator can read off the minimum achievable Stationary Generalization Error for a given system-cost budget without solving a nonconvex problem directly.
  • Algorithm 1, by matching representation distributions with symmetric KL penalties and adding Gaussian noise, empirically reduces the accuracy drop at drift points on Fashion-MNIST and CIFAR-10 across all three patterns.

Reading between the lines

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

  • Beyond the paper: the large-sample simplification is the load-bearing step; at the experimental size of $N=600$ samples per client, directly estimating $I(w_{\mathrm{cur}},Z)$ on the trained models would show whether the closed-form ordering still holds or is only asymptotic.
  • If the ordering is correct, an FL operator can allocate adaptation effort by pattern: under random drift, increasing the number of clients gives the largest payoff, while under gradual drift it gives the least.
  • The representation-space KL penalty in Algorithm 1 is a concrete, testable way to lower the KL terms in Eq. (13); varying the symmetry of the penalty (one-sided versus two-sided KL) would map which direction of distributional shift matters more.
  • The convex Pareto frontier in Theorem 2 suggests that online controllers could tune client processing capacities in real time without nonconvex solvers, an application the paper gestures at but does not develop.
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 / 5 minor

Summary. The paper proposes an information-theoretic analysis of federated learning under concept drift. Concept drift is modeled as a Markov chain over data distributions, and a Stationary Generalization Error G is defined as the steady-state expected weighted generalization error of the current global model on both the current and the next distribution. Theorem 1 states an upper bound on G in terms of mutual information I(w_cur, Z) and KL divergences between successive data distributions. The bound is then simplified for large sample sizes and sub-Gaussian losses, leading to closed-form expressions for periodic, gradual, and random drift patterns (Corollary 1) and an ordering result (Proposition 1). The paper also proposes a regularized ERM algorithm (KLMI-FedAvg), analyzes a performance-cost tradeoff via a Pareto front, and reports experiments on a Raspberry Pi testbed and on CIFAR-10, MNIST, Fashion-MNIST, and CIRCLE.

Significance. If the main theorem were fully established, the paper would provide a useful information-theoretic vocabulary for comparing drift patterns in FL and would connect the bound to an algorithmic regularizer and to a cost-performance tradeoff. The paper is largely non-parametric: the bound is a direct application of cited information-theoretic lemmas, with no hidden constants beyond the algorithm's hyperparameter gamma, and the testbed experiments are a concrete attempt at validation. However, the central theoretical claim is currently not proven for the losses used in the experiments, and the pattern-ordering results rest on an asymptotic simplification that is presented as a finite-sample bound. These gaps are load-bearing for the paper's main conclusions, so the contribution is significant only conditionally on repairing them.

major comments (3)
  1. [§IV-A and Appendix A, Eq. (12)] The proof of Theorem 1 invokes [29, Theorem 4] and, immediately after Eq. (33), states that 'we assume that the loss function is in the form of Bregman divergence.' This assumption is not stated in Theorem 1, Assumption 1, or the system model. It is load-bearing: the step from per-sample bounds (32) and (33) to the summed client bound (34) is exactly where the federated averaging structure is handled via [29, Theorem 4], and without the Bregman condition the change-of-measure argument does not produce the stated bound. The experiments in Sections VI-A and VI-B use cross-entropy losses, and the paper gives no argument that cross-entropy satisfies the Bregman condition. The theorem and the corollaries derived from it (Eqs. (15)–(17), Proposition 1) are therefore not established for the losses used in the paper. Please state the Bregman condition in Theorem 1 and either restrict the theoretical claims to Bregman losses or prove that the cross-entropy losses used in the experiments satisfy it.
  2. [§IV-B, Eq. (13)] Equation (13) is obtained by taking N to infinity and setting I(w_cur, Z) to zero, but it is introduced as an upper bound ('we can rewrite the bound as follows: G ≤ ...'). Because the omitted mutual-information terms are nonnegative, Eq. (13) is an asymptotic approximation of the bound in Theorem 1, not a finite-sample upper bound. Since Corollary 1 and Proposition 1 are derived from Eq. (13), the claimed ordering of periodic, gradual, and random patterns is only an asymptotic statement unless a finite-sample control of the mutual-information terms is provided. The experiments in Section VI-A use N=600 samples (Table I), so the verification in Figs. 7–10 does not directly test the proven finite-sample bound. Please restate Eq. (13) as an asymptotic result or add a finite-sample bound that accounts for the mutual-information terms.
  3. [§VI-A, Observation 3] Observation 3 states that, in contrast to the analytical result, no threshold p_th with B_p(p) > B_g(p) for p < p_th is observed in the CIFAR-10 experiments (Figs. 9 and 10). The paper attributes this to asymmetry of the KL divergence, but Proposition 1 is stated as a general result and Corollary 1 is derived under the symmetric Gaussian KL of Definition 3. The experiment therefore directly contradicts a stated theoretical claim. Please either make explicit the conditions (e.g., symmetric KL) under which Proposition 1 is claimed to hold, or revise the abstract and conclusion claims that the experiments corroborate the theoretical pattern ordering.
minor comments (5)
  1. [§I, Contributions] The phrase 'Desipite the noncovexity' should read 'Despite the nonconvexity'.
  2. [§IV-A, after Eq. (12)] The statement 'the function ψ*^{-1}(·) is concave' is asserted without proof and is not true for every Legendre dual arising in Theorem 1; the concavity claim should be stated under the sub-Gaussian or other specific ψ assumption used later.
  3. [§IV-B, Eq. (13)] If Lemma 1 and the sub-Gaussian assumption give ψ*^{-1}(x) = sqrt(2 r^2 x), then Eq. (13) appears to be missing a factor of sqrt(2); please check the consistency of the factor across Eqs. (13), (15)–(17), and Figure 5.
  4. [§III-A, Eq. (2)] The notation for the second sum in Eq. (2) and later in Theorem 1 is confusing: the upper limit is written as N while the normalization uses N_pre; please define N = N_cur + N_pre explicitly before Eq. (2) and keep the summation limits consistent.
  5. [§VI-B, Tables II–IV] The tables report a single accuracy value per method without error bars or multiple seeds; the claim that the proposed method 'significantly' outperforms baselines would be strengthened by reporting variance across runs.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the stationary-generalization-error bound is derived from external information-theoretic lemmas, with no fitted quantity renamed as a prediction; the only self-citation is confined to the cost model.

full rationale

Walking the derivation chain, Theorem 1 is obtained by substituting the population and empirical risk definitions into the stationary generalization error G and then applying cited external results: Assumption 1 follows [20,21], Lemma 1 follows [20, Lemma 2], the KL variational form follows [30], and the key per-sample bound in Appendix A follows [29, Theorem 4]. The MI and KL terms in Eq. (12) are genuine quantities, not free parameters, and no constant is fitted to data. The simplified bound in Eq. (13) is explicitly an asymptotic approximation obtained by taking I(w_cur,Z)->0 as N->infinity, so its finite-N status is a stated limitation rather than a circular step. Corollary 1 and Proposition 1 are algebraic consequences of Eq. (13), Definition 3, and the stated transition probabilities; they are not imports from the experiments. The one self-citation, [38], is used only for round duration and client cost in Section V-A and is not load-bearing for the information-theoretic bound. The experiments instantiate the same Markov-chain drift patterns that the theory assumes; this is a consistency check of the model, not a reduction of the prediction to its input, because the theorem is not derived from the experimental results. Separate correctness concerns do exist but are not circularity: the proof of Theorem 1 states in Appendix A that 'we assume that the loss function is in the form of Bregman divergence,' an assumption absent from the theorem statement and not verified for the cross-entropy losses used in the experiments, and Eq. (13) is an asymptotic simplification rather than a proven finite-sample upper bound. These gaps affect the validity of the claimed bound for the experimental settings but do not make the derivation circular.

Assumptions & free parameters 1 free parameters · 7 assumptions · 0 invented entities

The central bound depends on standard concentration assumptions on the loss (items 1-2), an unstated Bregman-divergence requirement (item 3), the ergodic Markov-chain model (item 4), and an asymptotic large-N step (item 5). The Gaussian KL computation is illustrative only (item 6). Algorithm 1 additionally assumes that representation-space KL is a proxy for data-space KL (item 7). The only tuned scalar is gamma.

free parameters (1)
  • gamma (gamma) = 0.001
    Regularization weight in Algorithm 1, chosen for experiments in Appendix F without sensitivity analysis.
assumptions (7)
  • domain assumption Assumption 1: cumulant generating function of loss is bounded by lambda and convex psi on [0,b)
    Standard in information-theoretic generalization bounds; not verified for the neural network losses used in experiments. Section IV-A.
  • domain assumption Loss function is sub-Gaussian with parameter r squared
    Used to set psi*^{-1}(x)=sqrt(2 r^2 x) in Section IV-B.
  • domain assumption Loss is a Bregman divergence (needed by reference [29], Theorem 4)
    Invoked in Appendix A to justify the per-client bound; absent from the statement of Theorem 1.
  • domain assumption The distribution sequence is a finite ergodic Markov chain that has reached steady state
    Definition 1 uses p(pi_pre) as the stationary distribution, requiring convergence and stationarity.
  • domain assumption N tends to infinity so mutual information I(w_cur,Z) tends to 0
    Used to derive Eq (13) and all Corollary 1 expressions; not valid as an upper bound for finite N without an error term. Section IV-B.
  • domain assumption Data distributions are multivariate Gaussian with equal covariance for KL computations
    Definition 3: not satisfied by image datasets in experiments, only by the theory's illustrative cases.
  • ad hoc to paper Representation-space KL divergence approximates data-distribution KL divergence
    Algorithm 1 estimates KL between learned representations, while Theorem 1 bounds KL between true data distributions; no proof connects the two. Section IV-C.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An Information-Theoretic Analysis for Federated Learning under Concept Drift." pith.science (2026). https://pith.science/paper/KYOGJCWF

@misc{pith2026250621036,
  author       = {Pith},
  title        = {Pith review of: An Information-Theoretic Analysis for Federated Learning under Concept Drift},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KYOGJCWF}},
  note         = {Machine review of arXiv:2506.21036}
}
read the original abstract

Recent studies in federated learning (FL) commonly train models on static datasets. However, real-world data often arrives as streams with shifting distributions, causing performance degradation known as concept drift. This paper analyzes FL performance under concept drift using information theory and proposes an algorithm to mitigate the performance degradation. We model concept drift as a Markov chain and introduce the \emph{Stationary Generalization Error} to assess a model's capability to capture characteristics of future unseen data. Its upper bound is derived using KL divergence and mutual information. We study three drift patterns (periodic, gradual, and random) and their impact on FL performance. Inspired by this, we propose an algorithm that regularizes the empirical risk minimization approach with KL divergence and mutual information, thereby enhancing long-term performance. We also explore the performance-cost tradeoff by identifying a Pareto front. To validate our approach, we build an FL testbed using Raspberry Pi4 devices. Experimental results corroborate with theoretical findings, confirming that drift patterns significantly affect performance. Our method consistently outperforms existing approaches for these three patterns, demonstrating its effectiveness in adapting concept drift in FL.

Figures

Figures reproduced from arXiv: 2506.21036 by the authors.

Figure 1
Figure 1. An illustration of an FL system under concept drift. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Users’ interest in news topics changes over time. [PITH_FULL_IMAGE:figures/full_fig_p001_2.png] view at source ↗
Figure 3
Figure 3. Inference accuracy decreases due to concept drift. Each time slot [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (15 more)
Figure 4
Figure 4. Figure 4: State machines for three patterns. (a) Periodic pattern. (b) Gradual pattern. (c) Random pattern. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: The upper bound of G changes over p. (a) D = 10, K = 10, α = 0.5, τ2 = 0.5. (b) D = 5, K = 100, α = 0.9, τ2 = 0.1. designed FL algorithms to effectively address the concept drift phenomenon, particularly in the context of the random pattern. Such an analytical result i…
Figure 7
Figure 7. Figure 7: CIRCLE: Impact on average test error with [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]
Figure 8
Figure 8. Figure 8: MNIST: Impact on average test error with [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]
Figure 9
Figure 9. Figure 9: CIFAR-10: Impact on average test error with [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 10
Figure 10. Figure 10: CIFAR-10: Impact on average test error with [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]
Figure 12
Figure 12. Figure 12: Accuracy curves for CIFAR-10 dataset (periodic pattern): (a) Training [PITH_FULL_IMAGE:figures/full_fig_p012_12.png]
Figure 11
Figure 11. Figure 11: Accuracy curves for Fashion-MNIST dataset (periodic pattern): [PITH_FULL_IMAGE:figures/full_fig_p012_11.png]
Figure 14
Figure 14. Figure 14: Pareto frontiers for gradual pattern. 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 C 10 20 30 40 B feasible domain K = 20 K = 50 K = 80 (a) K 0 10 20 30 40 50 C 10 20 30 40 B feasible domain D = 5 D = 10 D = 20 (b) D [PITH_FULL_IMAGE:figures/full_fig_p013_14.png]
Figure 15
Figure 15. Figure 15: Pareto frontiers for random pattern. Furthermore, we established an upper bound on G in terms of mutual information and KL divergence. To address per￾formance degradation caused by concept drift, we devel￾oped an algorithm to mitigate its effects. To explore the perfo…
Figure 17
Figure 17. Figure 17: Accuracy curves for CIFAR-10 dataset (random pattern): (a) Training [PITH_FULL_IMAGE:figures/full_fig_p016_17.png]
Figure 16
Figure 16. Figure 16: Accuracy curves for Fasion-MNIST dataset (random pattern): (a) [PITH_FULL_IMAGE:figures/full_fig_p016_16.png]
Figure 18
Figure 18. Figure 18: Accuracy curves for Fasion-MNIST dataset (gradual pattern): (a) [PITH_FULL_IMAGE:figures/full_fig_p016_18.png]
Figure 19
Figure 19. Figure 19: Accuracy curves for CIFAR-10 dataset (gradual pattern): (a) Training [PITH_FULL_IMAGE:figures/full_fig_p017_19.png]
Figure 20
Figure 20. Figure 20: Average test error changes over the distance between distributions. [PITH_FULL_IMAGE:figures/full_fig_p017_20.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 43 canonical work pages

  1. [1]

    Advances and open problems in federated learning,

    P. Kairouz, H. B. McMahan, B. Aventet al., “Advances and open problems in federated learning,”Found. Trends Mach. Learn., vol. 14, no. 1-2, pp. 1–210, 2021

    work page 2021

  2. [2]

    Incentive mechanism design for federated learning and unlearning,

    N. Ding, Z. Sun, E. Wei, and R. Berry, “Incentive mechanism design for federated learning and unlearning,” inProceedings of the 24th Inter- national Symposium on Theory, Algorithmic Foundations, and Protocol Design for Mobile Networks and Mobile Computing, MobiHoc, New York, NY , USA, October 2023

    work page 2023

  3. [3]

    Cache-enabled federated learning systems,

    Y . Liu, L. Su, C. Joe-Wong, S. Ioannidis, E. Yeh, and M. Siew, “Cache-enabled federated learning systems,” inProceedings of the 24th International Symposium on Theory, Algorithmic Foundations, and Protocol Design for Mobile Networks and Mobile Computing, MobiHoc, New York, NY , USA, October 2023

    work page 2023

  4. [4]

    Anarchic federated learning with delayed gradient averaging,

    D. Li and X. Gong, “Anarchic federated learning with delayed gradient averaging,” inProceedings of the 24th International Symposium on The- ory, Algorithmic Foundations, and Protocol Design for Mobile Networks and Mobile Computing, MobiHoc, New York, NY , USA, October 2023

    work page 2023

  5. [5]

    MIND: A large-scale dataset for news recommendation,

    F. Wu, Y . Qiao, J. Chen, C. Wu, T. Qi, J. Lian, D. Liu, X. Xie, J. Gao, W. Wu, and M. Zhou, “MIND: A large-scale dataset for news recommendation,” inProceedings of the 58th Annual Meeting of the Association for Computational Linguistics, ACL, Online, July 2020

    work page 2020

  6. [6]

    Learning under concept drift: A review,

    J. Lu, A. Liu, F. Dong, F. Gu, J. Gama, and G. Zhang, “Learning under concept drift: A review,”IEEE Trans. Knowl. Data Eng., vol. 31, no. 12, pp. 2346–2363, 2019

    work page 2019

  7. [7]

    A survey on concept drift adaptation,

    J. Gama, I. Zliobaite, A. Bifet, M. Pechenizkiy, and A. Bouchachia, “A survey on concept drift adaptation,”ACM Comput. Surv., vol. 46, no. 4, pp. 1–37, 2014

    work page 2014

  8. [8]

    Paired learners for concept drift,

    S. H. Bach and M. A. Maloof, “Paired learners for concept drift,” in Proceedings of the 8th IEEE International Conference on Data Mining, ICDM, Pisa, Italy, December 2008

    work page 2008

Show all 48 references
  1. [9]

    Learning from time-changing data with adaptive windowing,

    A. Bifet and R. Gavald `a, “Learning from time-changing data with adaptive windowing,” inProceedings of the 7th SIAM International Conference on Data Mining, Minneapolis, Minnesota, USA, April 2007

    2007

  2. [10]

    Mining time-changing data streams,

    G. Hulten, L. Spencer, and P. M. Domingos, “Mining time-changing data streams,” inProceedings of the 7th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco, CA, USA, August 2001

    2001

  3. [11]

    Accurate decision trees for mining high-speed data streams,

    J. Gama, R. Rocha, and P. Medas, “Accurate decision trees for mining high-speed data streams,” inProceedings of the 9th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Washington, DC, USA, August 2003

    2003

  4. [12]

    Dynamic extreme learning machine for data stream classification,

    S. Xu and J. Wang, “Dynamic extreme learning machine for data stream classification,”Neurocomputing, vol. 238, pp. 433–449, 2017

    2017

  5. [13]

    Federated continual learning with weighted inter-client transfer,

    J. Yoon, W. Jeong, G. Lee, E. Yang, and S. J. Hwang, “Federated continual learning with weighted inter-client transfer,” inProceedings of the 38th International Conference on Machine Learning, ICML, Virtual Event, July 2021

    2021

  6. [14]

    Fedhar: Semi-supervised online learning for personalized federated human activity recognition,

    H. Yu, Z. Chen, X. Zhang, X. Chen, F. Zhuang, H. Xiong, and X. Cheng, “Fedhar: Semi-supervised online learning for personalized federated human activity recognition,”IEEE Trans. Mob. Comput., vol. 22, no. 6, pp. 3318–3332, 2023

    2023

  7. [15]

    Federated learning under distributed concept drift,

    E. Jothimurugesan, K. Hsieh, J. Wang, G. Joshi, and P. B. Gibbons, “Federated learning under distributed concept drift,” inProceedings of the International Conference on Artificial Intelligence and Statistics, AISTATS, Palau de Congressos, Valencia, Spain, April 2023

    2023

  8. [16]

    Adap- tive federated learning in presence of concept drift,

    G. Canonaco, A. Bergamasco, A. Mongelluzzo, and M. Roveri, “Adap- tive federated learning in presence of concept drift,” inProceedings of the 2021 International Joint Conference on Neural Networks, IJCNN, Shenzhen, China, July 2021

    2021

  9. [17]

    Flash: Concept drift adaptation in federated learning,

    K. Panchal, S. Choudhary, S. Mitra, K. Mukherjee, S. Sarkhel, S. Mitra, and H. Guan, “Flash: Concept drift adaptation in federated learning,” inProceedings of the International Conference on Machine Learning, ICML, Honolulu, Hawaii, USA, July 2023

    2023

  10. [18]

    Shalev-Shwartz and S

    S. Shalev-Shwartz and S. Ben-David,Understanding Machine Learning - From Theory to Algorithms, 2014

    2014

  11. [19]

    Some pac-bayesian theorems,

    D. A. McAllester, “Some pac-bayesian theorems,”Mach. Learn., vol. 37, no. 3, pp. 355–363, 1999

    1999

  12. [20]

    Tightening mutual information- based bounds on generalization error,

    Y . Bu, S. Zou, and V . V . Veeravalli, “Tightening mutual information- based bounds on generalization error,”IEEE J. Sel. Areas Inf. Theory, vol. 1, no. 1, pp. 121–130, 2020

    2020

  13. [21]

    Information-theoretic analysis for transfer learning,

    X. Wu, J. H. Manton, U. Aickelin, and J. Zhu, “Information-theoretic analysis for transfer learning,” inProceedings of the IEEE International Symposium on Information Theory, ISIT, Los Angeles, CA, USA, June 2020

    2020

  14. [22]

    Principles of risk minimization for learning theory,

    V . Vapnik, “Principles of risk minimization for learning theory,” in Proceedings of the Advances in Neural Information Processing Systems 4, NIPS, Denver, Colorado, USA, December 1991

    1991

  15. [23]

    Decision tree classifier: a detailed survey,

    Priyanka and D. Kumar, “Decision tree classifier: a detailed survey,”Int. J. Inf. Decis. Sci., vol. 12, no. 3, pp. 246–269, 2020

    2020

  16. [24]

    Classifier clustering and feature alignment for federated learning under distributed concept drift,

    J. Chen, J. Xue, Y . Wang, Z. Liu, and L. Huang, “Classifier clustering and feature alignment for federated learning under distributed concept drift,” inProceedings of Advances in Neural Information Processing Systems 38: Annual Conference on Neural Information Processing Syst...

    2024

  17. [25]

    A multi-model approach for handling concept drifting data in federated learning,

    G. Yang, X. Chen, T. Zhang, and S. Wang, “A multi-model approach for handling concept drifting data in federated learning,” inWorkshop of Distributed Machine Learning and Unlearning for Sensor-Cloud Systems (DLS2) in the 20th International Conference on Mobility, Sensing and N...

    2024

  18. [26]

    A new approach to seasonal energy consumption forecasting using temporal convolutional networks,

    A. K. Shaikh, A. Nazir, N. Khalique, A. S. Shah, and N. Adhikari, “A new approach to seasonal energy consumption forecasting using temporal convolutional networks,”Results Eng., vol. 19, p. 101296, 2023. 14

    2023

  19. [27]

    Online consumer resilience during a pandemic: An exploratory study of e-commerce behavior before, during and after a covid-19 lockdown,

    C. Guthrie, S. Fosso-Wamba, and J. B. Arnaud, “Online consumer resilience during a pandemic: An exploratory study of e-commerce behavior before, during and after a covid-19 lockdown,”J. Retail. Consum. Serv., vol. 61, p. 102570, 2021

    2021

  20. [28]

    Detecting environmental change: science and society—perspectives on long-term research and monitoring in the 21st century,

    T. W. Parr, A. R. Sier, R. Battarbee, A. Mackay, and J. Burgess, “Detecting environmental change: science and society—perspectives on long-term research and monitoring in the 21st century,”Sci. Total Environ., vol. 310, no. 1-3, pp. 1–8, 2003

    2003

  21. [29]

    Improved information-theoretic generalization bounds for distributed, federated, and iterative learning,

    L. P. Barnes, A. Dytso, and H. V . Poor, “Improved information-theoretic generalization bounds for distributed, federated, and iterative learning,” Entropy, vol. 24, no. 9, p. 1178, 2022

    2022

  22. [30]

    Boucheron, G

    S. Boucheron, G. Lugosi, and P. Massart,Concentration Inequalities - A Nonasymptotic Theory of Independence, 2013

    2013

  23. [31]

    Stability and generalization,

    O. Bousquet and A. Elisseeff, “Stability and generalization,”J. Mach. Learn. Res., vol. 2, pp. 499–526, 2002

    2002

  24. [32]

    Information-theoretic analysis of generaliza- tion capability of learning algorithms,

    A. Xu and M. Raginsky, “Information-theoretic analysis of generaliza- tion capability of learning algorithms,” inProceedings of Advances in Neural Information Processing Systems 30: Annual Conference on Neu- ral Information Processing Systems, Long Beach, CA, USA, December 2017

    2017

  25. [33]

    On balancing bias and variance in unsupervised multi-source-free domain adaptation,

    M. Shen, Y . Bu, and G. W. Wornell, “On balancing bias and variance in unsupervised multi-source-free domain adaptation,” inProceedings of International Conference on Machine Learning, ICML 2023, Honolulu, Hawaii, USA, July 2023

    2023

  26. [34]

    How does distribution matching help domain generalization: An information- theoretic analysis,

    Y . Dong, T. Gong, H. Chen, S. Song, W. Zhang, and C. Li, “How does distribution matching help domain generalization: An information- theoretic analysis,”IEEE Trans. Inf. Theory, vol. 71, no. 3, pp. 2028– 2053, 2025

    2025

  27. [35]

    KL guided domain adaptation,

    A. T. Nguyen, T. Tran, Y . Gal, P. H. S. Torr, and A. G. Baydin, “KL guided domain adaptation,” inProceedings of the 10th International Conference on Learning Representations, ICLR, Virtual Event, April 2022

    2022

  28. [36]

    Communication-efficient learning of deep networks from decentralized data,

    B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas, “Communication-efficient learning of deep networks from decentralized data,” inProceedings of the 20th International Conference on Artificial Intelligence and Statistics, AISTATS, Fort Lauderdale, FL, USA, April 2017

    2017

  29. [37]

    Federated optimization in heterogeneous networks,

    T. Li, A. K. Sahu, M. Zaheer, M. Sanjabi, A. Talwalkar, and V . Smith, “Federated optimization in heterogeneous networks,” inProceedings of the 3rd Conference on Machine Learning and Systems, MLSys, Austin, TX, USA, March 2020

    2020

  30. [38]

    A blockchain-empowered incentive mechanism for cross-silo federated learning,

    M. Tang, F. Peng, and V . W. Wong, “A blockchain-empowered incentive mechanism for cross-silo federated learning,”IEEE Trans. Mob. Com- put., vol. 23, no. 10, pp. 9240–9253, 2024

    2024

  31. [39]

    Federated learning over wireless networks: Optimization model design and analysis,

    N. H. Tran, W. Bao, A. Y . Zomaya, M. N. H. Nguyen, and C. S. Hong, “Federated learning over wireless networks: Optimization model design and analysis,” inProceedings of the 2019 IEEE Conference on Computer Communications, INFOCOM, Paris, France, April 2019

    2019

  32. [40]

    Learning with drift detection,

    J. Gama, P. Medas, G. Castillo, and P. P. Rodrigues, “Learning with drift detection,” inProceedings of the 17th Brazilian Symposium on Artificial Intelligence, SBIA, S ˜ao Luis, Maranh ˜ao, Brazil, September 2004

    2004

  33. [41]

    The MNIST database,

    Y . LeCun, C. Cortes, and C. J.C. Burges, “The MNIST database,” http: //yann.lecun.com/exdb/mnist/, accessed: 2024-04-15

    2024

  34. [42]

    Learning multiple layers of features from tiny images,

    A. Krizhevsky, G. Hintonet al., “Learning multiple layers of features from tiny images,” 2009

    2009

  35. [43]

    Non-iid data and continual learning processes in federated learning: A long road ahead,

    M. F. Criado, F. E. Casado, R. Iglesias, C. V . Regueiro, and S. Barro, “Non-iid data and continual learning processes in federated learning: A long road ahead,”Inf. Fusion, vol. 88, pp. 263–280, 2022

    2022

  36. [44]

    Measuring the effects of non-identical data distribution for federated visual classification,

    T. H. Hsu, H. Qi, and M. Brown, “Measuring the effects of non-identical data distribution for federated visual classification,”arXiv preprint, vol. abs/1909.06335, 2019

    1909 arXiv

  37. [45]

    Speedtest global index,

    SPEEDTEST, “Speedtest global index,” https://www.speedtest.net/ global-index, accessed: 2024-04-15

    2024

  38. [46]

    Cloud services pricing,

    Microsoft Azure, “Cloud services pricing,” https://azure.microsoft.com/ en-us/pricing/details/cloud-services, accessed: 2024-04-15

    2024

  39. [47]

    En- ergy consumption in mobile phones: a measurement study and im- plications for network applications,

    N. Balasubramanian, A. Balasubramanian, and A. Venkataramani, “En- ergy consumption in mobile phones: a measurement study and im- plications for network applications,” inProceedings of the 9th ACM SIGCOMM Internet Measurement Conference, IMC, Chicago, Illinois, USA, November 2009

    2009

  40. [48]

    Electricity prices in Canada 2020,

    E. Hub, “Electricity prices in Canada 2020,” https://www.energyhub.org/ electricity-prices/, accessed: 2024-04-15. APPENDIXA PROOF OFTHEOREM1 Theorem 1(Upper Bound of the Stationary Generalization Error).The Stationary Generalization Error is bounded as: G≤ 1 K 2 X πpre,πcur,π...

    2020

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.