Pith. sign in

REVIEW 4 major objections 4 minor 22 references

Quantum-Evolutionary Neural Networks for Multi-Agent Federated Learning

T0 review · 4 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a federated learning loop with quantum-style sine activations, Gaussian mutation-selection, and privacy noise converges in expectation while matching centralized accuracy on standard image benchmarks.

desk verdict The paper's convergence guarantee is disproved by a simple counterexample, the privacy claim is uncheckable, and the experiments are illustrative; desk reject, though the mutation-in-FL idea is worth noting. read the letter →

arxiv 2505.15836 v1 pith:M6NOBWIY submitted 2025-05-16 cs.NE cs.AIcs.LG

Aarav Lala , Kalyan Cherukuri This is my paper · ORCID
classification cs.NEcs.AIcs.LG
keywords quantumneuralnetworksevolutionaryalgorithmsfederatedlearningprivacy-preservingAImulti-agentsystemsdifferentialprivacysineactivation
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

The paper proposes Quantum-Evolutionary Neural Networks (QE-NN) trained in a federated loop: each client mutates the global model into $K$ Gaussian-perturbed variants, fine-tunes each, keeps the best, adds Gaussian noise, and sends the result to a server that averages all client updates. The central claim is that this loop converges in expectation to a stationary point of the global loss under smoothness, bounded-gradient, and bounded-variance assumptions, and that the per-round update satisfies $(\epsilon,\delta)$-differential privacy. The appeal is a decentralized multi-agent training scheme that never shares raw data yet still reaches accuracy comparable to a centralized baseline on MNIST, CIFAR-10, and CIFAR-100. If true, this would give privacy-sensitive applications (autonomous systems, smart cities, healthcare) a way to keep models adaptive without central data pooling.

What carries the argument

The load-bearing objects are three. First, the QuantumLayer maps activations through $z = \sin(Wz + \phi)$ with trainable phase shifts $\phi$, which the paper reads as emulating superposition and functional entanglement across stacked layers. Second, the local evolutionary step draws $K$ independent Gaussian perturbations around the global model, fine-tunes each variant for $E$ local epochs, and keeps the variant with lowest local loss; this 'min over $K$' is the improvement oracle behind Eq. (11) and the source of the expected loss reduction in Eq. (9). Third, the privacy mechanism adds Gaussian noise $\delta_i$ to each selected model before the server averages all $N$ models, with the noise scale $\sigma_p$ controlling the claimed per-round DP guarantee.

What would settle it

On a small convex problem (e.g., federated logistic regression on MNIST), record the per-round global loss and gradient norms for $K=1$ and $K=10$; if the running average of squared gradient norms does not decrease toward zero under any constant $\eta$, or if the decrease rate contradicts Eq. (9), the convergence claim fails. Alternatively, compute the $\ell^2$ sensitivity of the aggregated update $(1/N)\sum_i(\theta^*_i+\delta_i)$ over neighboring datasets; if the sensitivity exceeds $\Delta$, the per-round DP bound $\epsilon = \Delta^2/(2\sigma_p^2)$ is violated.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a periodic, quantum-inspired activation ($z = \sin(Wz + \phi)$) combined with local evolutionary selection and privacy-preserving federated averaging forms a trainable framework with two formal guarantees: Eq. (10) asserts that the average squared gradient norm over training rounds goes to zero, and Eq. (12) asserts that each round is $(\epsilon,\delta)$-differentially private with $\epsilon = \Delta^2/(2\sigma_p^2)$. The experiments in Figure 4 show accuracy, F1, and loss curves for the federated QE-NN staying close to the baseline across MNIST, CIFAR-10, and CIFAR-100, with the synthetic study reporting the global model stabilizing around 0.97 accuracy.

Load-bearing premise

The proof assumes without derivation that the whole mutation-select-and-average step is equivalent to one stochastic gradient step on the global loss with a well-defined effective learning rate $\eta$, even though $\eta$ is never defined and the selection over $K$ variants is never shown to be unbiased or to have bounded bias.

Editorial extensions

If this is right

  • A deployed QE-FL system can keep raw data on each agent and still train a shared model, since only noise-perturbed selected variants are transmitted.
  • The per-round DP guarantee gives an explicit knob: increasing $\sigma_p$ lowers $\epsilon$ at the cost of slower convergence (via $\sigma_p^2$ in Eq. (9)).
  • The evolutionary selection step guarantees that unless a client is already at a local minimum, it finds a strictly better variant with positive probability, acting as a local improvement oracle under non-convex losses.
  • Accuracy, F1, and loss on MNIST, CIFAR-10, and CIFAR-100 stay close to the centralized baseline, so privacy-preserving distributed training does not require a large accuracy sacrifice on these benchmarks.
  • The sine phase-shift architecture's periodicity yields a richer representation landscape that the evolutionary search can explore.

Reading between the lines

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

  • The bias introduced by the min over $K$ variants is never bounded; a natural extension is to characterize selection bias as a function of $K$ and $\sigma$, which would turn Eq. (9) into a quantitative convergence bound rather than an existence claim.
  • The paper gives per-round DP only; composing $R$ rounds would multiply $\epsilon$ unless advanced composition or a privacy accountant is applied, so a total-privacy version of Eq. (12) is an immediate open step.
  • The periodicity of $\sin(Wz+\phi)$ creates infinitely many stationary points at phase shifts separated by $2\pi$; the evolutionary selection might be the mechanism that escapes poor phase basins, a testable hypothesis by comparing $K=1$ vs large $K$ on a phase-sensitive task.
  • The empirical comparison in Figure 4 is against a 'baseline,' but the paper does not specify whether that baseline is centralized QE-NN or standard FedAvg; rerunning with both would clarify what the evolutionary-quantum components actually add over plain federated averaging.
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

4 major / 4 minor

Summary. The paper introduces Quantum-Evolutionary Neural Networks (QE-NN), a federated learning framework in which each client generates K Gaussian-perturbed copies of the global model, fine-tunes them locally, selects the variant with lowest local loss, adds Gaussian noise for privacy, and sends the result to a server for federated averaging. The authors claim a convergence guarantee (Eq. (9)-(10)), a differential-privacy guarantee (Eq. (12)), and support these claims with a synthetic experiment and comparisons on MNIST, CIFAR10, and CIFAR100. The central technical claims are stated as theorems but are not derived from the stated assumptions, and the experimental section does not provide the details needed to validate the results.

Significance. If the convergence and privacy guarantees were established rigorously, the combination of evolutionary selection, quantum-inspired activations, and federated averaging could be of interest to the federated learning community as a gradient-free local-search mechanism. However, the paper's main mathematical claims are asserted without proof, the privacy formula is incorrect as stated, and the experiments do not compare against a proper federated baseline or report sufficient statistical detail. The paper also introduces terms such as 'functional entanglement' without formal definition. At present the significance is limited by the lack of supported guarantees and inadequate empirical validation.

major comments (4)
  1. [4.1, Eq. (9)] The key convergence inequality is asserted directly after the smoothness and bounded-variance assumptions, with no derivation and no explicit definition of eta or C. The min-over-K selection step in Eq. (5) creates a bias that cannot be absorbed into a fixed effective learning rate. For a concrete counterexample, take N=1, K=2, sigma_p=0, no local SGD, and L(theta)=0.5||theta||^2, which satisfies Assumptions 1-3. For scalar theta >> sigma, the selected perturbation is essentially the smaller of two Gaussian draws, so the expected update is theta - c*sigma*sign(theta) for a constant c>0, not theta - eta*grad L(theta). Hence the term -eta||grad L||^2 in Eq. (9) cannot be produced, and Eq. (10) does not follow. The authors need to either provide a full derivation bounding the selection bias or explicitly restrict the setting in which the claim holds.
  2. [4.3, Eq. (12)] The claimed differential-privacy guarantee is not correct. The Gaussian mechanism gives epsilon = Delta * sqrt(2 log(1.25/delta)) / sigma_p (in the standard formulation), not epsilon = Delta^2/(2 sigma_p^2). Moreover, the sensitivity Delta of the clients' model updates is never defined or bounded for the QE-NN update rule, and the composition across R rounds is not addressed. Consequently, the statement that each round satisfies (epsilon, delta)-DP is unsupported. The authors should derive the privacy guarantee from a precise sensitivity analysis of the actual update mechanism, including the selection step and the fine-tuning.
  3. [5, Table 1 and Figure 3] The synthetic experiment reports the accuracy of the best of 10 mutated models (Table 1) on the same data used for selection, which is a selection-bias artifact; Figure 3 then shows the accuracy of the global model over rounds, but no baseline (e.g., standard FedAvg or a centralized model) is provided for comparison. The text states that these trends 'validate the theoretical advantages of our evolutionary approach,' but with selection on the training set and no comparator, the experiment does not support this conclusion.
  4. [5, Figure 4] The experimental section does not report essential hyperparameters nor the exact protocol for the MNIST, CIFAR10, and CIFAR100 comparisons: number of clients, degree of non-IID sharding, local epochs E, learning rate eta, mutation standard deviation sigma, privacy noise sigma_p, number of variants K, or the aggregation rule. Without these details and without error bars or multiple runs, the reported comparisons cannot be reproduced or evaluated statistically.
minor comments (4)
  1. [1, Abstract and Introduction] The abstract and introduction make broad claims of a 'breakthrough' and 'pushing the boundaries of AI,' which are not supported by the presented evidence; suggest a more measured tone.
  2. [2.2] There is a typographical error in 'non-differential non-convex, or noisy objective landscapes'; the intended term is likely 'non-differentiable.'
  3. [3.2 and 4.2] The term 'functional entanglement' is used but never defined formally; the sinusoidal activation in Eq. (3) is a standard nonlinearity and does not by itself implement quantum entanglement, so the analogy should be stated more carefully.
  4. [5, Figure 3/Figure 4] The text refers to 'Figure 4c shows that the accuracy of the global model consistently improves,' but Figure 4 is described as a comparison across MNIST, CIFAR10, and CIFAR100; the figure numbering and cross-references should be harmonized.

Circularity Check

3 steps flagged · score 4.0 of 10

Partial circularity: the 'evolutionary benefit,' the synthetic confirmation, and the quantum layer reduce to the algorithm's own selection/definition; the main convergence theorem is a bare assertion rather than a derivation.

  1. self definitional [Section 4.3, Eq. (11)]
    "Assuming Gaussian perturbations are sufficiently expressive to explore local minima, the probability of selecting a strictly better variant than θ is strictly positive unless θ is already a local minimum: P[∃k s.t. L_i(θ+ϵ_k^(i)) < L_i(θ)] > 0 if ∇L_i(θ) ≠ 0"

    This statement is a direct restatement of the selection construction in Eqs. (4)-(5): the client samples K Gaussian perturbations and defines θ*_i as the argmin of the local loss. If ∇L_i(θ)≠0, then by L-smoothness some small step in the negative-gradient direction decreases L_i, and the Gaussian perturbation has positive density in that descent cone. The event therefore has positive probability by construction, so presenting it as an 'evolutionary benefit' is the algorithm's own selection rule restated as a theorem. It also does not supply the quantitative bias bound needed for Eq. (9); positivity alone cannot justify a fixed effective learning rate η.

  2. fitted input called prediction [Section 5, Table 1 and the following paragraph]
    "As shown in Table 1, mutation M1 achieved the highest accuracy at 0.973. Most of the models stayed around the 0.94-0.96 range, with some outliers. This distribution confirms the strength of our mutation-selection mechanism in improving local model quality."

    The table reports exactly the quantities used by the selection rule: Algorithm 1 (lines 11-13) evaluates each mutated variant's loss on the local dataset and keeps the best one. For a fixed labeled set, lowest loss is the same criterion as highest accuracy, so reporting the best of ten mutations as 'confirming the strength of our mutation-selection mechanism' is the selection objective back-formed as evidence. No baseline accuracy of the unmutated global model on the same evaluation set is provided, so 'improvement' is not measured independently; the confirmation reduces to the argmin/argmax used to choose M1.

1 more flagged steps
  1. renaming known result [Sections 3.2 and 4.2, Eq. (3)]
    "This periodic architecture enables the network to capture complex, nonlinear interactions between features, inspired by quantum superposition and entanglement. ... We emulate this phenomenon through stacked QuantumLayer blocks in the QENN architecture. Each layer’s output depends not only on the input features but also on the globally learned phase shift, effectively creating a shared encoding across the model’s internal layers. This can be viewed as a form of functional entanglement."

    The architecture defined by Eq. (3) is a linear layer followed by a phase-shifted sine activation, z=sin(Wz+ϕ), a classical periodic nonlinearity with a trainable phase offset. The paper never constructs a quantum state, a unitary, or a measurement map; the terms 'superposition' and 'functional entanglement' are asserted by analogy ('inspired by,' 'can be viewed as'). Labeling this standard activation as quantum-inspired renames a known classical component rather than deriving a result from quantum mechanics, so the claimed quantum benefit is carried by the naming rather than by new mathematical content.

full rationale

We found no self-citation or imported-uniqueness circularity: the bibliography is external and none of the cited works is used as a load-bearing substitute for the paper's own derivation. The strongest claimed result, Eq. (9), is not a circular reduction: the paper never defines η and never bounds the bias introduced by the min-over-K selection, so the inequality is an unsupported assertion rather than a consequence of Assumptions 1-3. Similarly, Eq. (12) invokes a standard (if imprecisely stated) Gaussian privacy mechanism rather than a self-referential construction. The genuine circular/definitional issues are localized: Eq. (11) restates the selection construction, the synthetic validation in Table 1 reports the selected-best accuracy on the same evaluation set used for selection, and the quantum content of Eq. (3) is a renamed sine activation. These issues undermine parts of the claimed novelty, but the external MNIST/CIFAR comparisons and the privacy mechanism give the paper some independent empirical content, so the overall circularity is partial rather than total.

Assumptions & free parameters 5 free parameters · 5 assumptions · 2 invented entities

The framework's theoretical claims rest on standard smoothness and boundedness assumptions plus two ad hoc assumptions about mutation expressiveness and bounded sensitivity that are not established. The 'quantum-inspired' layer is a periodic activation, and the 'entanglement' is shared phase parameters, so the quantum vocabulary adds no independent constraint. The ledger makes clear that the central derivation is not self-contained.

free parameters (5)
  • mutation standard deviation sigma = not reported
    Algorithm 1 line 7; controls exploration in evolutionary selection and appears in the convergence bound in Eq. (9).
  • privacy noise standard deviation sigma_p = not reported
    Algorithm 1 line 16; controls the claimed differential privacy guarantee in Eq. (12) and affects convergence.
  • number of variants K = 10 in the synthetic experiment
    K variants per client; the experimental choice is not justified or varied.
  • local epochs E and learning rate eta = not reported
    Used in the local SGD step in Algorithm 1 lines 8-10 and in the claimed effective learning rate in Eq. (9).
  • privacy regularization coefficient lambda = not reported
    Appears in Eq. (8) but is absent from Algorithm 1, so its optimization role is undefined.
assumptions (5)
  • domain assumption Each local loss L_i(theta) is L-smooth.
    Section 4.1 Assumption 1; standard but unverified for the neural network losses used in the experiments.
  • domain assumption Gradients of local losses are uniformly bounded by G.
    Section 4.1 Assumption 2; not generally true for neural networks without gradient clipping, which is not specified.
  • domain assumption Mutation and privacy noises are zero mean with bounded variance.
    Section 4.1 Assumption 3; consistent with the Gaussian choices, but the paper does not verify the variance bounds.
  • ad hoc to paper Gaussian perturbations are sufficiently expressive to explore local minima.
    Stated in Section 4.3 to justify Eq. (11); no coverage argument or convergence rate is given.
  • ad hoc to paper The sensitivity Delta of model updates is bounded.
    Required for Eq. (12), but no bound on Delta is derived after K mutations and local fine-tuning.
invented entities (2)
  • functional entanglement
    purpose: Claims that stacked QuantumLayer blocks create a quantum-like coupling through shared learned phase shifts across layers.
    Section 4.2 describes it as a relabeling of weight and phase sharing in a standard feed-forward network; there is no quantum state, no entanglement measure, and no falsifiable prediction.
  • QuantumLayer
    purpose: Custom layer used in QE-NN; the paper says it simulates superposition via x -> sin(Wx + phi).
    This is a sine-activated linear layer, not a quantum circuit, and it provides no independent evidence of quantum advantage.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Quantum-Evolutionary Neural Networks for Multi-Agent Federated Learning." pith.science (2026). https://pith.science/paper/M6NOBWIY

@misc{pith2026250515836,
  author       = {Pith},
  title        = {Pith review of: Quantum-Evolutionary Neural Networks for Multi-Agent Federated Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/M6NOBWIY}},
  note         = {Machine review of arXiv:2505.15836}
}
read the original abstract

As artificial intelligence continues to drive innovation in complex, decentralized environments, the need for scalable, adaptive, and privacy-preserving decision-making systems has become critical. This paper introduces a novel framework combining quantum-inspired neural networks with evolutionary algorithms to optimize real-time decision-making in multi-agent systems (MAS). The proposed Quantum-Evolutionary Neural Network (QE-NN) leverages quantum computing principles -- such as quantum superposition and entanglement -- to enhance learning speed and decision accuracy, while integrating evolutionary optimization to continually refine agent behaviors in dynamic, uncertain environments. By utilizing federated learning, QE-NN ensures privacy preservation, enabling decentralized agents to collaborate without sharing sensitive data. The framework is designed to allow agents to adapt in real-time to their environments, optimizing decision-making processes for applications in areas such as autonomous systems, smart cities, and healthcare. This research represents a breakthrough in merging quantum computing, evolutionary optimization, and privacy-preserving techniques to solve complex problems in multi-agent decision-making systems, pushing the boundaries of AI in real-world, privacy-sensitive applications.

Figures

Figures reproduced from arXiv: 2505.15836 by the authors.

Figure 1
Figure 1. From superposition to entanglement: A single qubit [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Quantum-Evolutionary Federated Learning Pipeline [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Accuracy of Global Model on Synthetic Dataset over Training Rounds [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Performance comparison across MNIST, CIFAR10, and CIFAR100 for [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

22 extracted references · 20 canonical work pages

  1. [12]

    Ortiz Marrero, C., Wiebe, N., Furches, J., Ragone, M.: Quantum neural networks: Issues, training, and applications. Tech. Rep. PNNL-XXXX, Pacific Northwest National Laboratory (2023), https://www.pnnl.gov/publications/quantum-neural- networks-issues-training-and-applications

    work page 2023

  2. [20]

    Journal of Machine Learning Research (JMLR)24, 1–24 (2023), https://www.jmlr.org/papers/volume24/2023/2023 evolutionary.pdf

    Yu, J., Kim, S.J., Lee, J., Choi, J.: Evolutionary search and gradient-based update for efficient federated learning. Journal of Machine Learning Research (JMLR)24, 1–24 (2023), https://www.jmlr.org/papers/volume24/2023/2023 evolutionary.pdf

    work page 2023

  3. [1]

    Sensors23(6), 2957 (2023)

    Cagnoni, S., Mazzoni, A., Susi, G., Bechini, A.: A federated learning-inspired evo- lutionary algorithm: Application to glucose prediction. Sensors23(6), 2957 (2023). https://doi.org/10.3390/s23062957

    work page doi:10.3390/s23062957 2023

  4. [2]

    Evolutionary Intelligence 16(3), 1033–1044 (2023)

    Chai, Z.y., Yang, C.d., Li, Y.l.: Communication efficiency optimization in federated learning based on multi-objective evolutionary algorithm. Evolutionary Intelligence 16(3), 1033–1044 (2023)

    work page 2023

  5. [3]

    Quantum Information Processing22(5), 223 (2023)

    Friedrich, L., Maziero, J.: Evolution strategies: application in hybrid quantum- classical neural networks. Quantum Information Processing22(5), 223 (2023). https://doi.org/10.1007/s11128-023-03876-8

    work page doi:10.1007/s11128-023-03876-8 2023

  6. [4]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

    Gao, J., Zhang, K., Zhang, S., Liu, A.: Feddc: Federated learning with non-iid data via local drift decoupling and correction. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). pp. 10102– 10112 (2022)

    work page 2022

  7. [5]

    In: International Conference on Machine Learning (ICML) (2017), https://arxiv.org/abs/1703.01041

    Jaderberg, M., Czarnecki, W.M., Parisotto, E., Sifre, L., Lillicrap, T., Leibo, J.Z., Kavukcuoglu, K.: Population based training of neural net- works. In: International Conference on Machine Learning (ICML) (2017), https://arxiv.org/abs/1703.01041

    arXiv 2017

  8. [6]

    Multiagent and Grid Systems20(1), 1–20 (2024)

    Jenefa, A., Vidhya, K., Taurshia, A., Naveen, V., Kuriakose, B., Vijula, V.: Enhancing distributed agent environments with quantum multi-agent systems and protocols. Multiagent and Grid Systems20(1), 1–20 (2024). https://doi.org/10.3233/MGS-230127

    work page doi:10.3233/mgs-230127 2024

Show all 22 references
  1. [7]

    Physical Review Research1(3), 033063 (2019)

    Killoran, N., Bromley, T.R., Arrazola, J.M., Schuld, M., Quesada, N., Lloyd, S.: Continuous-variable quantum neural networks. Physical Review Research1(3), 033063 (2019)

    2019

  2. [8]

    In: Pro- ceedings of the 2020 International Conference on Learning Representations (ICLR) (2020), https://openreview.net/forum?id=rkfg5k4twB 12 A

    Liu, Y., Chen, J., Zhang, Y., Li, L., Yu, T., Tan, Z., Li, K., Wang, Y.: Fedsel: A federated evolutionary selection method for robust model aggregation. In: Pro- ceedings of the 2020 International Conference on Learning Representations (ICLR) (2020), https://openreview.net/for...

    2020

  3. [9]

    In: Proceedings of the 20th International Conference on Artificial Intelligence and Statistics (AISTATS)

    McMahan, H.B., Moore, E., Ramage, D., Hampson, S., et al.: Communication- efficient learning of deep networks from decentralized data. In: Proceedings of the 20th International Conference on Artificial Intelligence and Statistics (AISTATS). pp. 1273–1282. PMLR (2017)

    2017

  4. [10]

    In: Proceedings of the 32nd Inter- national Conference on Neural Information Processing Systems (NeurIPS) (2019), https://arxiv.org/abs/1901.10125

    Miikkulainen, R., Elazari, N., Finkel, H., Leong, J., Liu, C.H., Strube, J., Van Der Meer, D.: Evolving deep neural networks. In: Proceedings of the 32nd Inter- national Conference on Neural Information Processing Systems (NeurIPS) (2019), https://arxiv.org/abs/1901.10125

    2019 arXiv

  5. [11]

    ´A., Dahi, Z.A., Chicano, F., Luque, G., Alba, E.: Optimising commu- nication overhead in federated learning using nsga-ii

    Morell, J. ´A., Dahi, Z.A., Chicano, F., Luque, G., Alba, E.: Optimising commu- nication overhead in federated learning using nsga-ii. In: Proceedings of the 2022 Genetic and Evolutionary Computation Conference Companion. pp. 1–2. ACM (2022)

    2022

  6. [13]

    Advances in Neural Information Processing Systems36, 62428–62441 (2023)

    Rahimi, M.M., Bhatti, H.I., Park, Y., Kousar, H., Moon, J.: Evofed: leveraging evolutionary strategies for communication-efficient federated learning. Advances in Neural Information Processing Systems36, 62428–62441 (2023)

    2023

  7. [14]

    In: Proceedings of the AAAI Con- ference on Artificial Intelligence

    Sattler, F., Wiedemann, K.R., M¨ uller, N., Samek, W.: Robust and communication- efficient federated learning from non-iid data. In: Proceedings of the AAAI Con- ference on Artificial Intelligence. vol. 34, pp. 6140–6149 (2020)

    2020

  8. [15]

    The journal of physical chemistry letters 12(43), 10654–10662 (2021)

    Secor, M., Soudackov, A.V., Hammes-Schiffer, S.: Artificial neural networks as propagators in quantum dynamics. The journal of physical chemistry letters 12(43), 10654–10662 (2021)

    2021

  9. [16]

    In: Advances in Neural Information Processing Systems (NeurIPS)

    Shao, Y., Wang, Z., Lin, Q., Yang, J., Gong, S.: Coded robust and secure feder- ated learning with secret sharing. In: Advances in Neural Information Processing Systems (NeurIPS). vol. 35, pp. 31215–31228 (2022)

    2022

  10. [17]

    Quantum Machine Intelligence6(2), 1–13 (2024)

    Singh, U., Goldberg, A.Z., Heshami, K.: Coherent feed-forward quantum neural network. Quantum Machine Intelligence6(2), 1–13 (2024)

    2024

  11. [18]

    In: Pham, D.N., Theeramunkong, T., Governatori, G., Liu, F

    Wei, B., Li, J., Liu, Y., Wang, W.: Federated learning for non-iid data: From theory to algorithm. In: Pham, D.N., Theeramunkong, T., Governatori, G., Liu, F. (eds.) PRICAI 2021: Trends in Artificial Intelligence, Lecture Notes in Computer Science, vol. 13031, pp. 34–46. Sprin...

    2021 doi

  12. [19]

    Physical Review A102(5), 052421 (2020)

    Yan, S., Qi, H., Cui, W.: Nonlinear quantum neuron: A fundamental building block for quantum neural networks. Physical Review A102(5), 052421 (2020)

    2020

  13. [21]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Yun, W.J., Park, J., Kim, J.: Quantum multi-agent meta reinforcement learning. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 37, pp. 10892–10900. AAAI Press (2023). https://doi.org/10.1609/aaai.v37i9.26313

    2023 doi

  14. [22]

    In: Das- gupta, S., Mandt, S., Li, Y

    Zakerinia, H., Talaei, S., Nadiradze, G., Alistarh, D.: Communication- efficient federated learning with data and client heterogeneity. In: Das- gupta, S., Mandt, S., Li, Y. (eds.) Proceedings of The 27th International Conference on Artificial Intelligence and Statistics. Proc...

    2024

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.