Rehearsal-Free Continual Federated Learning with Synergistic Synaptic Intelligence

Haozhao Wang; Ruixuan Li; Tianzhe Xiao; Yichen Li; Yining Qi; Yuying Wang

REVIEW 3 major objections 4 minor 3 cited by

Rehearsal-Free Continual Federated Learning with Synergistic Synaptic Intelligence

T0 review · 3 major / 4 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 FedSSI's rehearsal-free regularizer beats CFL baselines by up to 12.47% in final accuracy.

desk verdict A useful empirical CFL paper with a plausible PSM mechanism that is not cleanly isolated from SI-penalty tuning; worth refereeing, but the authors should report missing hyperparameters and trim the theory claims. read the letter →

arxiv 2412.13779 v3 pith:GHL5TTC5 submitted 2024-12-18 cs.LG cs.DC

Yichen Li , Yuying Wang , Haozhao Wang , Yining Qi , Tianzhe Xiao , Ruixuan Li This is my paper · ORCID

classification cs.LGcs.DC

keywords continualfederatedlearningsynapticintelligenceregularizationnon-IIDdatacatastrophicforgettingrehearsal-freepersonalizedsurrogatemodelclass-incremental

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

Continual federated learning (CFL) systems that let clients learn from streaming tasks usually fight catastrophic forgetting by replaying cached or synthetic past samples, which costs memory and can violate privacy. This paper makes the case that regularization-based continual learning can work in the federated setting instead, and identifies synaptic intelligence (SI) as the strongest regularization baseline under IID data but one that collapses under non-IID data. The authors propose FedSSI, which adds a personalized surrogate model (PSM) at each client so that the parameter-importance scores SI relies on are computed with both local and global information. The claim is that this simple, rehearsal-free modification lets a regularized CFL method beat state-of-the-art rehearsal and non-rehearsal baselines across six datasets and two incremental scenarios, with gains up to 12.47% in final accuracy.

What carries the argument

The load-bearing object is the Personalized Surrogate Model (PSM), a per-client auxiliary model that is never used for classification. Before each new task arrives, the client updates the PSM for a few local iterations on its previous-task samples, with an added proximal pull toward the last global model; the scalar $q(\lambda) = (1-\lambda)/(2\lambda)$ controls the trade-off, with $\lambda\to 0$ making the PSM converge to the global model and larger $\lambda$ keeping it local. The gradient of the PSM's loss, rather than the gradient of the target model, is integrated over the training trajectory to compute the synaptic-intelligence contributions $s^k_{l,i}$ in Eq. (6), which then feed the importance scores $\Omega^k_{l,i}$ that penalize movement of old-task weights when training the new task in Eq. (2). This mechanism is what transfers global knowledge into the regularization penalty without any rehearsal, sample caching, or extra communication.

What would settle it

Measure the cosine similarity between the per-parameter importance vector $\Omega$ computed by FedSSI and the local-only SI importance vector of FL+SI on a strongly non-IID CIFAR10 split ($\alpha=0.1$); if the two vectors are nearly identical while final accuracy still differs by several points, the global pull through the PSM is not the mechanism producing the reported gains.

Watch

Extended reading notes

Core claim

FedSSI's central claim is that the failure of regularization-based CFL under data heterogeneity is not a failure of synaptic intelligence itself but of where its importance estimates come from. In vanilla FL+SI, each client computes the surrogate-loss contributions $s^k_{l,i}$ using only its local model and local data, so the importance weights $\Omega^k_{l,i}$ protect weights that matter locally, which can be misaligned with what the global model needs under non-IID data. FedSSI replaces that local-only computation with a Personalized Surrogate Model (PSM) $v^k_{t-1}$ trained on the client's previous-task data while being pulled toward the received global model $w^{t-1}$ by a proximal term $q(\lambda)(v - w)$, with $q(\lambda) = (1-\lambda)/(2\lambda)$. The PSM's gradient path is then used in place of the local model's to accumulate SI importance scores, so the surrogate loss in Eq. (2) penalizes changes to weights that matter for both the local and the global data distributions. The paper reports that with this change, FedSSI achieves the best final and average accuracy in all tested cases, including gains of up to 12.47% in final accuracy, and stays ahead of baselines as data heterogeneity $\alpha$ varies.

Load-bearing premise

The proximal pull toward the global model in Eq. (5) makes each client's personalized surrogate model carry enough global knowledge that the resulting importance scores protect the right weights under non-IID data; if that pull adds no useful global signal, FedSSI reduces to plain FL+SI and its reported gains lack a mechanism.

Editorial extensions

If this is right

Rehearsal-based CFL methods (such as Re-Fed, FedCIL, and GLFC) can be outperformed by a regularization-only method, so memory buffers and generative replay are not necessary for state-of-the-art CFL accuracy.
The PSM training costs about one fortieth of the per-task training budget and requires no extra communication, keeping the rehearsal-free benefit cheap in practice.
The $\lambda$ knob gives practitioners a way to respond to the degree of data heterogeneity: decreasing $\lambda$ (more global pull) is the direction that recovers accuracy under stronger non-IID splits, a trend confirmed in Table 3 across three datasets.
FedSSI works for both class-incremental and domain-incremental tasks without needing task boundaries at inference, unlike FOT and FedWeIT, which the paper had to modify to automate task-ID inference.

Reading between the lines

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

Extension: the same 'compute importance on a globally-pulled surrogate' recipe should transfer to other regularization-based continual learners that use per-weight importance, such as EWC-style Fisher estimates; if it does, FedSSI is a template rather than a single algorithm.
Extension: the method implicitly assumes the client still has the just-completed task's data when the PSM is updated; if data must be discarded the moment a task ends, PSM training would need to run online during the task itself, a variant the paper does not test.
Extension: the $\lambda$–$\alpha$ trend in Table 3 suggests an automatic scheduling rule for $\lambda$ could be fitted from the Dirichlet concentration parameter, removing the per-dataset tuning burden that the paper leaves manual.

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.

Desk Editor's Note

read the letter

Bottom line first: this is a solid empirical paper for the continual-federated-learning subfield, not a conceptual breakthrough. The new piece is modest and real — FedSSI takes the personalized surrogate model from the authors' own Re-Fed line and uses it to compute synaptic-intelligence importances under a global anchor, instead of using it to select replay samples. The ingredients are all known (SI, proximal personalization, Re-Fed's PSM), but the combination is new and the motivation is sensible: plain SI's surrogate loss reflects only the local distribution, which misaligns with the global objective when client data is non-IID. Figures 1–2 document that failure cleanly.

The experiments are the strength. Six datasets, class-IL and domain-IL, a dozen baselines, plus ablations over heterogeneity level, number of tasks, scalability, and bandwidth. FedSSI wins across the board, and the margins over FL+SI are consistent rather than cherry-picked. Table 3's λ–α trend is a genuine empirical pattern, even if the paper overstates what Proposition 1 proves about it.

Now the soft spots, in proportion. The α notation collision is the most visible sin — α is the SI penalty in Eq. (2) and the Dirichlet heterogeneity parameter. The SI penalty strength is never reported for FedSSI or FL+SI, which is the real gap: the mechanistic claim that the PSM's importance estimates, rather than a better-tuned SI penalty, explain the Table 2 gains is not fully isolated. That said, the stress-test speculation that penalty tuning alone could produce the gaps does not fully land — FedSSI also beats replay baselines, and the gains are consistent across datasets. The cheap fix is to report the penalty strengths and add an ablation where FL+SI gets the same tuning care.

The theory section is the weakest part. Proposition 1 is imported from Hanzely–Richtárik and only shows λ trades local vs. global in the PSM objective; it does not establish the positive α–λ correlation the text claims. Theorem 1 is a self-citation from Li et al. 2024a with no proof or independent check. The paper would be more credible calling this a plausibility argument.

Who it's for: anyone building rehearsal-free CFL methods or extending regularization-based CL to federated settings will want this as a reference point. My recommendation: send it to review. The revisions needed are reporting and framing, not redoing the experiments.

Referee Report

3 major / 4 minor

Summary. The paper proposes FedSSI, a rehearsal-free regularization method for continual federated learning (CFL) that adapts Synaptic Intelligence (SI) to non-IID data. The key idea is a personalized surrogate model (PSM) per client, trained on the client's previous local task with a proximal pull toward the global model; the PSM's parameter contributions are used to compute SI importance weights that penalize changes to important weights when learning new tasks. The authors first show that standard regularization methods, especially FL+SI, work well under IID client data but degrade under non-IID partitions, and then argue that FedSSI's PSM restores performance. Experiments on six datasets under Class-IL and Domain-IL scenarios compare FedSSI against FedAvg, FedProx, regularization baselines (FL+LwF, FL+EWC, FL+OGD, FL+SI), and CFL baselines (Re-Fed, FedCIL, GLFC, FOT, FedWeIT), reporting final and average accuracy, sensitivity to the Dirichlet heterogeneity parameter, communication-round efficiency, and resource overhead. The paper claims up to 12.47% improvement in final accuracy over state-of-the-art methods and includes a short analytical section on the PSM.

Significance. If the central claim holds, FedSSI would be a valuable rehearsal-free CFL method that avoids memory and privacy costs of replay while addressing data heterogeneity. The experimental protocol is broad: six datasets, two incremental scenarios, multiple data-heterogeneity levels, and many baselines, with both final and average accuracy reported. The paper also reports communication-efficiency trade-offs and computational overhead, which is useful for practitioners. However, the significance is conditional: the unreported SI penalty strength and the lack of an ablation isolating the PSM mechanism prevent the current results from being uniquely attributed to the proposed method. The analytical section is largely imported from prior work (Hanzely & Richtarik 2020; Li et al. 2024a) and does not establish the paper's main claimed α–λ relationship.

major comments (3)

[Section 5.1, Eq. (2), Table 5] The SI regularization coefficient α in Eq. (2) is never reported for any method, while the same symbol α is used in Section 5.1 as the Dirichlet heterogeneity parameter. This makes the comparison between FedSSI and FL+SI in Table 2 confounded: the gains (e.g., +3.26 on CIFAR10, +4.69 on Tiny-ImageNet) could be due to a more favorable SI penalty strength rather than the PSM mechanism. Please report the exact SI penalty coefficients used for FL+SI and FedSSI for every dataset, and include an ablation in which the SI penalty is held fixed while the PSM is removed or replaced.
[Section 4.3, Proposition 1, Theorem 1] The claimed positive correlation between the heterogeneity parameter α and the PSM balance parameter λ is not established by the provided analysis. Proposition 1 only shows that as λ→0 the PSM in Eq. (7) approaches the global model for a strongly convex objective, and it does not involve the Dirichlet α or the SI penalty in Eq. (2). Theorem 1 is imported from Li et al. 2024a and assumes the global model converges to the optimum at rate g(t), which is precisely the condition at issue in non-IID CFL, and no proof or independent verification is provided. Table 3 gives empirical evidence for only three λ values. Thus the paper's theoretical support for the central mechanism is insufficient.
[Section 4.2, Algorithm 1] The paper does not provide an ablation that isolates the contribution of the PSM over ordinary SI. The PSM in Eq. (5) is trained only on local previous-task data with a proximal pull to the global model, and the paper asserts that this yields importance estimates reflecting both local and global distributions. Since FedSSI adds both the PSM and a second tuned hyperparameter λ, a comparison against plain SI with the same SI penalty strength but with importance computed from the global model, or against FedSSI with the PSM replaced by a global-mean importance estimate, is needed to show that the PSM, rather than regularization-strength tuning or the extra λ, drives the observed improvements.

minor comments (4)

[Table 2] The column header "CIFAI100" appears to be a typo for "CIFAR100".
[Section 5.2, Resource Consumption] The sentence "In the next version, we will consider using a smaller network model..." refers to future work and should be removed or the corresponding experiment should be included; published manuscripts should not defer a verification step to a later version.
[Section 4.2] The notation α is used for both the SI penalty coefficient in Eq. (2) and the Dirichlet heterogeneity parameter in the experiments; these are different quantities and should be renamed to avoid confusion.
[Section 4.2, discussion after Eq. (5)] The statement "Denote that α refers to the degree of data heterogeneity and when α has a higher value, indicating a trend towards homogeneity in distribution" is ambiguous: a larger Dirichlet α means more homogeneous data, so it would be clearer to say that α controls heterogeneity and larger values correspond to more IID settings.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the central claim is empirical and benchmark-based; the only notable issue is a self-cited convergence theorem in the theory section, which is a theory gap rather than a circular derivation.

full rationale

The paper's central claim is empirical: FedSSI outperforms regularization and rehearsal baselines across six datasets (Table 2, Figure 3). These comparisons are self-contained against external benchmarks, and the method is fully specified by Eqs. (2)-(7), so the main result does not reduce to a fit or to a definition. I find no step where a 'prediction' is identical by construction to an input. The SI surrogate loss in Eq. (2) uses importance estimates computed from PSM gradients via Eqs. (5)-(6); Proposition 1 in Section 4.3 is inherited from Hanzely-Richtarik and only characterizes how the PSM objective in Eq. (7) interpolates between the local and global models as lambda changes. It does not assume the paper's performance conclusion. The one substantial self-citation is Theorem 1 in Section 4.3, attributed to the authors' own Li et al. 2024a and stated without proof: 'Here we introduce Theorem 1 proposed by (Li et al., 2024a), which is stated by our proposed personalized surrogate model. Based on it, we ensure the convergence of the PSM and the effectiveness of FedSSI can be proved along by Proposition 1.' This is an overreach: the theorem assumes w_t converges to the optimal model, and convergence of the PSM does not by itself establish final-accuracy gains under non-IID CFL. That is a logical gap in the theoretical narrative, not circularity, because the benchmark evidence does not depend on the theorem. Two confounds deserve note but are also not circularity: the symbol alpha is reused for the SI penalty coefficient in Eq. (2) and for the Dirichlet heterogeneity parameter in Sections 4.2 and 5.1, and the SI penalty strength used for FedSSI and FL+SI is never reported, so part of the observed gap over FL+SI could reflect regularization-strength tuning rather than the PSM mechanism alone. These are correctness and ablation concerns, not cases where the derivation reduces to its own inputs. Overall score 2.

Assumptions & free parameters 3 free parameters · 4 assumptions · 1 invented entities

The central method is a heuristic assembled from two existing ingredients: the SI importance measure and a personalized proximal model similar to those in the authors' earlier Re-Fed and SR-FDIL work. No proof shows that the PSM's local-global mixture is what improves forgetting; the only formal statements are a borrowed strong-convexity proposition and a self-cited convergence theorem. The empirical comparisons depend on a tuned lambda and on access to previous-task data for the PSM update.

free parameters (3)

lambda (global/local balance in PSM update) = 0.2 to 0.8 in Table 3; selection per dataset and alpha not fully specified in Table 2
Controls q(lambda)=(1-lambda)/(2*lambda) in Eq. (5), the strength of the pull toward the global model; the authors report best values per heterogeneity level.
SI regularization coefficient alpha in Eq. (2) = not reported
Scales the surrogate loss in L_total; its value is never listed in the experimental setup, and the symbol alpha also denotes Dirichlet heterogeneity, causing ambiguity.
PSM local update steps s = not reported
Algorithm 1 updates the PSM for 's local iterations'; the paper says this costs 1/40 of training but does not state s, which affects the computed importance path.

assumptions (4)

domain assumption Objective f(v_k) is strongly convex (Definition 1)
Proposition 1 borrows Hanzely-Richtarik analysis under strong convexity, but the actual ResNet18 training objective is non-convex, so the analysis does not apply directly.
standard math Global model w_t converges to w_hat with rate g(t) (Theorem 1)
Convergence is cited from Li et al. 2024a, the authors' own prior work, and used to claim PSM convergence; no proof appears in this paper.
ad hoc to paper PSM trained on previous local task data with a proximal pull to the global model encodes global knowledge
This is the core mechanism (Eqs. 5 and 6); no theorem establishes that the resulting per-weight contributions align with the global data distribution.
domain assumption Clients can access previous-task samples during the task transition to train the PSM
Algorithm 1 requires local previous-task samples to update the PSM; if data from completed tasks is already deleted for privacy, this step cannot run as written.

invented entities (1)

Personalized Surrogate Model (PSM)
purpose: Auxiliary per-client model used only to compute per-weight importance for synaptic intelligence; anchors local updates to the global model via Eq. (5).
The PSM is an algorithmic construct with no falsifiable prediction outside the FedSSI pipeline; its only evidence is end-task accuracy.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Rehearsal-Free Continual Federated Learning with Synergistic Synaptic Intelligence." pith.science (2026). https://pith.science/paper/GHL5TTC5

@misc{pith2026241213779,
  author       = {Pith},
  title        = {Pith review of: Rehearsal-Free Continual Federated Learning with Synergistic Synaptic Intelligence},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GHL5TTC5}},
  note         = {Machine review of arXiv:2412.13779}
}

read the original abstract

Continual Federated Learning (CFL) allows distributed devices to collaboratively learn novel concepts from continuously shifting training data while avoiding knowledge forgetting of previously seen tasks. To tackle this challenge, most current CFL approaches rely on extensive rehearsal of previous data. Despite effectiveness, rehearsal comes at a cost to memory, and it may also violate data privacy. Considering these, we seek to apply regularization techniques to CFL by considering their cost-efficient properties that do not require sample caching or rehearsal. Specifically, we first apply traditional regularization techniques to CFL and observe that existing regularization techniques, especially synaptic intelligence, can achieve promising results under homogeneous data distribution but fail when the data is heterogeneous. Based on this observation, we propose a simple yet effective regularization algorithm for CFL named FedSSI, which tailors the synaptic intelligence for the CFL with heterogeneous data settings. FedSSI can not only reduce computational overhead without rehearsal but also address the data heterogeneity issue. Extensive experiments show that FedSSI achieves superior performance compared to state-of-the-art methods.

Figures

Figures reproduced from arXiv: 2412.13779 by the authors.

**Figure 2.** Performance comparisons of aforementioned methods on CIFAR10 and Digit10 datasets with Non-IID data. the surrogate loss for the previous tasks. w t k,i represents the i-th parameter of local model w t k in client k and w t−1 represents the optimal weights in the (t-1)-th timestamp, estimated based on its importance for previous tasks. Client k utilizes Ω t k,i to measure the importance of the i-th parameter of the… view at source ↗

**Figure 3.** Performance w.r.t data heterogeneity α for four datasets. WeIT (Yoon et al., 2021); three models designed for continual federated learning with data rehearsal: Re-Fed (Li et al., 2024a), FedCIL (Qi et al., 2023), and GLFC (Dong et al., 2022); and four custom methods combining traditional CL techniques with the FedAvg algorithm: FL+LwF (Li & Hoiem, 2017), FL+EWC (Kirkpatrick et al., 2017), FL+OGD (Farajtabar et al.… view at source ↗

**Figure 4.** Performance w.r.t number of incremental tasks n for two class-incremental datasets D. Additional Results In this section, we first provide the experiments about training time cost and quantitative analysis. Then, we conduct experiments on scalability and bandwidth constraints to validate the effectiveness of our method. Finally, we provide more details about the experiment results of each task. We record the test ac… view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

SUM: Unified Geometric Surgery on Spatio-Temporal Adaptation Vectors for Federated Class Incremental Learning
cs.LG 2026-07 conditional novelty 6.0 of 10

SUM projects client and task adaptation vectors to remove directional interference during server aggregation, improving federated class-incremental learning accuracy without client-side changes.
Task-Agnostic Federated Continual Learning via Replay-Free Gradient Projection
cs.LG 2025-09 conditional novelty 6.0 of 10

FedProTIP combines gradient projection onto orthogonal complements of prior-task subspaces with subspace-relevance task-identity prediction to achieve replay-free, task-agnostic federated continual learning.
Unleashing the Power of Continual Learning on Non-Centralized Devices: A Survey
cs.LG 2024-12 conditional novelty 4.0 of 10

A review of non-centralized continual learning that taxonomizes data-, model-, and device-level methods and benchmarks twelve federated continual learning methods on six datasets.

Reference graph

Works this paper leans on

21 extracted references · 7 canonical work pages · cited by 3 Pith papers

[8]

Re-evaluating continual learning scenarios: A categorization and case for strong baselines

Hsu, Y .-C., Liu, Y .-C., and Kira, Z. Re-evaluating continual learning scenarios: A categorization and case for strong baselines. ArXiv, abs/1810.12488,

arXiv
[10]

Towards efficient replay in federated incremental learning

Li, Y ., Li, Q., Wang, H., Li, R., Zhong, W., and Zhang, G. Towards efficient replay in federated incremental learning. The Thirty-Fifth IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition (CVPR 2024), Seattle, USA, June 17-21, 2024a. Li, Y ., Xu, W., Qi, Y ., Wang, H., Li, R., and Guo, S. Sr- fdil: Synergistic replay for federated domain-inc...

arXiv 2024
[11]

Federated recommendation with additive personalization, 2024c

Li, Z., Long, G., and Zhou, T. Federated recommendation with additive personalization, 2024c. URL https:// arxiv.org/abs/2301.09109. Liu, B., Lv, N., Guo, Y ., and Li, Y . Recent advances on federated learning: A systematic survey. Neu- rocomputing, 597:128019,

arXiv
[12]

doi: https://doi.org/10.1016/j.neucom.2024.128019

ISSN 0925-2312. doi: https://doi.org/10.1016/j.neucom.2024.128019. URL https://www.sciencedirect.com/ science/article/pii/S0925231224007902. Liu, C., Qu, X., Wang, J., and Xiao, J. Fedet: a communication-efficient federated class-incremental learning framework based on enhanced transformer. arXiv preprint arXiv:2306.15347,

work page arXiv 2024
[13]

Long, M., Cao, Y ., Wang, J., and Jordan, M. I. Learn- ing transferable features with deep adaptation networks. ArXiv, abs/1502.02791,

arXiv
[17]

Adapting visual category models to new domains

Saenko, K., Kulis, B., Fritz, M., and Darrell, T. Adapting visual category models to new domains. In Computer Vision–ECCV 2010: 11th European Conference on Com- puter Vision, Heraklion, Crete, Greece, September 5-11, 2010, Proceedings, Part IV 11 , pp. 213–226. Springer,

work page 2010
[19]

Dafkd: Domain-aware federated knowledge distillation

Wang, H., Li, Y ., Xu, W., Li, R., Zhan, Y ., and Zeng, Z. Dafkd: Domain-aware federated knowledge distillation. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, pp. 20412–20421, 2023a. Wang, H., Xu, H., Li, Y ., Xu, Y ., Li, R., and Zhang, T. Fed- cda: Federated learning with cross-rounds divergence- aware aggregati...

work page 2024
[20]

Sola: Continual learning with second-order loss approximation

Yin, D., Farajtabar, M., and Li, A. Sola: Continual learning with second-order loss approximation. ArXiv, abs/2006.10974,

arXiv 2006

Show all 21 references

[21]

Semantic drift compensation for class-incremental learning

Yu, L., Twardowski, B., Liu, X., Herranz, L., Wang, K., Cheng, Y ., Jui, S., and van de Weijer, J. Semantic drift compensation for class-incremental learning. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 6980–6989,

2020
[22]

Datasets Class-Incremental Task Dataset: New classes are incrementally introduced over time

11 FedSSI: Rehearsal-Free Continual Federated Learning with Synergistic Synaptic Intelligence A. Datasets Class-Incremental Task Dataset: New classes are incrementally introduced over time. The dataset starts with a subset of classes, and new classes are added in subsequent st...

2010
[1994]

Communication-efficient on-device machine learning: Federated distillation and augmentation under non-iid private data

Jeong, E., Oh, S., Kim, H., Park, J., Bennis, M., and Kim, S. Communication-efficient on-device machine learning: Federated distillation and augmentation under non-iid private data. CoRR, abs/1811.11479, 2018a. Jeong, E., Oh, S., Kim, H., Park, J., Bennis, M., and Kim, S.- L. ...

arXiv
[1996]

Domain- incremental continual learning for mitigating bias in facial expression and action unit recognition

Churamani, N., Kara, O., and Gunes, H. Domain- incremental continual learning for mitigating bias in facial expression and action unit recognition. ArXiv, abs/2103.08637,

arXiv
[2010]

10 FedSSI: Rehearsal-Free Continual Federated Learning with Synergistic Synaptic Intelligence van de Ven, G. M. and Tolias, A. S. Three scenarios for continual learning. ArXiv, abs/1904.07734,

1904 arXiv
[2015]

Con- tinual federated learning based on knowledge distilla- tion

Ma, Y ., Xie, Z., Wang, J., Chen, K., and Shou, L. Con- tinual federated learning based on knowledge distilla- tion. In Raedt, L. D. (ed.), Proceedings of the Thirty- First International Joint Conference on Artificial Intel- ligence, IJCAI-22 , pp. 2182–2188. International Joi...

doi
[2017]

Domain-adversarial training of neural networks

Ganin, Y ., Ustinova, E., Ajakan, H., Germain, P., Larochelle, H., Laviolette, F., Marchand, M., and Lempitsky, V . Domain-adversarial training of neural networks. The journal of machine learning research, 17(1):2096–2030,
[2019]

S., Blundell, C., Zwols, Y ., Ha, D

Fernando, C., Banarse, D. S., Blundell, C., Zwols, Y ., Ha, D. R., Rusu, A. A., Pritzel, A., and Wierstra, D. Path- net: Evolution channels gradient descent in super neural networks. ArXiv, abs/1701.08734,

arXiv
[2020]

He, K., Zhang, X., Ren, S., and Sun, J

URL https://arxiv.org/abs/2002.05516. He, K., Zhang, X., Ren, S., and Sun, J. Deep residual learn- ing for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 770–778,

2002 arXiv
[2021]

Em- nist: Extending mnist to handwritten letters

Cohen, G., Afshar, S., Tapson, J., and Van Schaik, A. Em- nist: Extending mnist to handwritten letters. In 2017 in- ternational joint conference on neural networks (IJCNN), pp. 2921–2926. IEEE,

2017
[2022]

Better generative re- play for continual federated learning

Qi, D., Zhao, H., and Li, S. Better generative re- play for continual federated learning. arXiv preprint arXiv:2302.13001,
[2023]

Rebuffi, S.-A., Kolesnikov, A., Sperl, G., and Lampert, C. H. icarl: Incremental classifier and representation learning. In Proceedings of the IEEE conference on Computer Vision and Pattern Recognition, pp. 2001–2010,

2001
[2024]

F., Yaldiz, D

Bakman, Y . F., Yaldiz, D. N., Ezzeldin, Y . H., and Aves- timehr, S. Federated orthogonal training: Mitigating global catastrophic forgetting in continual federated learn- ing. arXiv preprint arXiv:2309.01289,

arXiv

Pith tools

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

[1] [8]

Re-evaluating continual learning scenarios: A categorization and case for strong baselines

Hsu, Y .-C., Liu, Y .-C., and Kira, Z. Re-evaluating continual learning scenarios: A categorization and case for strong baselines. ArXiv, abs/1810.12488,

arXiv

[2] [10]

Towards efficient replay in federated incremental learning

Li, Y ., Li, Q., Wang, H., Li, R., Zhong, W., and Zhang, G. Towards efficient replay in federated incremental learning. The Thirty-Fifth IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition (CVPR 2024), Seattle, USA, June 17-21, 2024a. Li, Y ., Xu, W., Qi, Y ., Wang, H., Li, R., and Guo, S. Sr- fdil: Synergistic replay for federated domain-inc...

arXiv 2024

[3] [11]

Federated recommendation with additive personalization, 2024c

Li, Z., Long, G., and Zhou, T. Federated recommendation with additive personalization, 2024c. URL https:// arxiv.org/abs/2301.09109. Liu, B., Lv, N., Guo, Y ., and Li, Y . Recent advances on federated learning: A systematic survey. Neu- rocomputing, 597:128019,

arXiv

[4] [12]

doi: https://doi.org/10.1016/j.neucom.2024.128019

ISSN 0925-2312. doi: https://doi.org/10.1016/j.neucom.2024.128019. URL https://www.sciencedirect.com/ science/article/pii/S0925231224007902. Liu, C., Qu, X., Wang, J., and Xiao, J. Fedet: a communication-efficient federated class-incremental learning framework based on enhanced transformer. arXiv preprint arXiv:2306.15347,

work page arXiv 2024

[5] [13]

Long, M., Cao, Y ., Wang, J., and Jordan, M. I. Learn- ing transferable features with deep adaptation networks. ArXiv, abs/1502.02791,

arXiv

[6] [17]

Adapting visual category models to new domains

Saenko, K., Kulis, B., Fritz, M., and Darrell, T. Adapting visual category models to new domains. In Computer Vision–ECCV 2010: 11th European Conference on Com- puter Vision, Heraklion, Crete, Greece, September 5-11, 2010, Proceedings, Part IV 11 , pp. 213–226. Springer,

work page 2010

[7] [19]

Dafkd: Domain-aware federated knowledge distillation

Wang, H., Li, Y ., Xu, W., Li, R., Zhan, Y ., and Zeng, Z. Dafkd: Domain-aware federated knowledge distillation. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, pp. 20412–20421, 2023a. Wang, H., Xu, H., Li, Y ., Xu, Y ., Li, R., and Zhang, T. Fed- cda: Federated learning with cross-rounds divergence- aware aggregati...

work page 2024

[8] [20]

Sola: Continual learning with second-order loss approximation

Yin, D., Farajtabar, M., and Li, A. Sola: Continual learning with second-order loss approximation. ArXiv, abs/2006.10974,

arXiv 2006