Pith. sign in

REVIEW 4 major objections 5 minor 32 references

Online Decentralized Federated Multi-task Learning With Trustworthiness in Cyber-Physical Systems

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

Pith's one-line read An online decentralized multi-task federated learner proves O(√T) regret for honest clients even when Byzantine attackers outnumber them, given physical-layer trust probabilities that separate honest from malicious signals.

desk verdict Real problem, plausible approach, but the main theorem has no proof and the experiments assume the key assumption. read the letter →

arxiv 2509.00992 v1 pith:ALCIUBUD submitted 2025-08-31 cs.LG cs.AI

classification cs.LGcs.AI
keywords onlinefederatedlearningdecentralizedmulti-taskByzantineresiliencetrustprobabilitycyber-physicalsystemsregretboundconstraintviolation
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 claims that federated learning can stay resilient even when Byzantine attackers outnumber honest clients—a regime where standard Byzantine-robust methods fail—if clients exploit cyber-physical side information. The authors design an online decentralized multi-task learning algorithm in which each honest client computes a running trust score for every neighbor from a probability derived from physical signals (e.g., received signal strength), then updates its personalized model using only neighbors whose trust score is nonnegative. They prove that under a mean-separation assumption on those trust probabilities, the honest clients' cumulative regret is O(√T) and their constraint violation is O(T^(3/4)), matching the best-known rates for the Byzantine-free case. Simulations with 30 Byzantine and 15 honest clients show performance close to a Byzantine-free benchmark. If correct, this is the first algorithm with sublinear-regret guarantees for a Byzantine majority in online decentralized federated multi-task learning.

What carries the argument

The running trust score βvu(t)=Σ_{k=0}^{t-1}(αvu(k)−0.5), whose sign selects the trusted neighbor set Nh_v(t)={u∈Nv | βvu(t)≥0} at each round. This sign-threshold filter is the mechanism that lets the algorithm discard Byzantine updates; the regret analysis rests on the Hoeffding-style decay of misclassification probability from Lemma 1, applied to the trusted-set gradients in the regularized Lagrangian primal-dual update (Algorithm 1).

What would settle it

Run Algorithm 1 with trust probabilities whose means are reversed or equal for honest and Byzantine neighbors (e.g., both E[α]=0.5), so Assumption 1(ii) fails; measure the honest clients' time-average regret over a long horizon. If regret becomes linear, the separation is load-bearing; if it stays sublinear, the bound does not depend on the claimed trust margin. A complementary test: mount a signal-strength spoofing attack that makes Byzantine transmissions mimic honest ones and check whether the trusted-neighbor filter allows malicious updates through.

Watch

Extended reading notes

Core claim

The central claim is that malicious clients can be outvoted, not by robust statistics on the model updates, but by physical evidence about the sender. Each honest client v maintains βvu(t), the accumulated difference between the observed trust probability αvu(t) and 0.5 for each neighbor u. By a concentration argument, βvu(t) is eventually positive for honest neighbors and negative for Byzantine neighbors, so the sign of βvu(t) defines the trusted set Nh_v(t) that enters the saddle-point update. The algorithm then runs projected primal-descent (on the local loss plus similarity constraints to trusted neighbors) and dual-ascent (on the constraint violations) restricted to that trusted set. Th

Load-bearing premise

For every honest client, the expected trust score minus 0.5 must be nonnegative for honest neighbors and negative for Byzantine neighbors; if an attacker can spoof the physical signal so that this separation fails, the trusted set admits Byzantine updates and the regret bound collapses.

Editorial extensions

If this is right

  • Honest clients with a weak but nonnegative trust margin still enjoy sublinear regret; the bound contains O(√T) terms and O(T^(3/4)) constraint violation.
  • The algorithm needs no central server and no consensus; each client builds its own trusted neighbor set from its own physical observations.
  • Time-varying data distributions are handled directly, so the method applies to autonomous and wireless settings where data drift is the norm.
  • The long-run average constraint satisfaction means model-similarity (personalization) constraints are met even under a Byzantine majority.
  • Simulation with 30 Byzantine and 15 honest clients tracks the Byzantine-free benchmark, suggesting the filter removes nearly all malicious influence after a short burn-in.

Reading between the lines

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

  • The paper leaves the breakdown threshold unquantified: its own future-work section says it will seek a bound on how far the Byzantine majority can grow before the algorithm fails, implying the O(√T) guarantee may hold only up to some majority size not characterized here.
  • Because the Theorem's proof is omitted from this text, the exact conditions under which the modification of the existing proof goes through cannot be checked from the paper alone; a reader cannot verify whether additional assumptions beyond Assumption 1 are needed.
  • The trust model is generic: any side information satisfying the same mean separation (reputation scores, protocol compliance, spatial fingerprints) would plug into the same algorithm, since only the scalar αvu(t) enters the update.
  • The early-round behavior is not covered by the regret bound: Lemma 1 guarantees misclassification decays exponentially but provides no finite time Tf, so the trusted set in the first rounds can still admit Byzantine updates before the filter stabilizes.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The manuscript proposes an online decentralized federated multi-task learning algorithm (Algorithm 1) that combines regularized Lagrangian optimization with a trust-score filter based on cyber-physical observations. The trust score α_vu(t) is used to build a time-varying trusted-neighbor set N_v^h(t) by thresholding β_vu(t), the cumulative excess of α_vu(t) over 0.5. The central claim is Theorem 1 in Section IV: under Assumption 1, honest clients running Algorithm 1 achieve static regret O(√T) and constraint violation O(T^{3/4}) even when Byzantine clients form a majority. The paper supports this with a simulation of 15 honest and 30 Byzantine clients on a fully connected graph, using online logistic regression and trust values drawn with means 0.55 (honest) and 0.45 (Byzantine). The empirical claim is that performance is close to the Byzantine-free baseline.

Significance. If the theorem and the trust-separation premise held, the paper would address an important open gap: Byzantine-resilient federated learning under a Byzantine majority. Combining online decentralized multi-task learning with physical-layer trust is an appealing direction, and the authors correctly note that standard statistical-robustness filters fail when more than half of the clients are Byzantine. However, the current manuscript does not provide enough evidence for the central claim. The proof of the main theorem is omitted entirely, the stated regret bound contains an ill-defined term, and the trust-separation assumption is used as a simulation input rather than validated or even stress-tested. The paper is best viewed as a promising extended abstract whose core technical contribution remains unverified.

major comments (4)
  1. [Section IV, Theorem] The proof of Theorem 1 is stated as 'omitted due to space constraints' and described only as a modification of Theorem 1 in [32]. For a journal submission, the central regret and constraint-violation guarantees cannot be accepted on a citation alone. The proof must be supplied, either in full or in an appendix, with clear derivations of Eqs. (22) and (23). Without this, the main contribution is unverifiable.
  2. [Section IV, Eq. (22)] The regret bound is syntactically ill-defined. The term β_vu appears in the denominator inside the expression `(1 + 1/(ζ β_vu))/(1 - (1+ζ)β_vu)` with no lower bound on β_vu, no definition of the aggregate `β` used in the interval `ζ ∈ (0, 1/β - 1)`, and no handling of β_vu = 0. As written, the claimed O(√T) bound is not a mathematically valid upper bound. This suggests the proof (or its transcription) is not carefully worked out.
  3. [Assumption 1(ii) and Algorithm 1, Step 9] Assumption 1(ii) is load-bearing: the trusted-neighbor filter β_vu(t) ≥ 0 separates honest from Byzantine neighbors only if E[α_vu] - 0.5 is nonnegative for honest neighbors and negative for Byzantine neighbors. The manuscript assumes α_vu(t) is 'known from the physicality' and cites [23], [24], but the simulation in Section V samples α directly from distributions with means 0.55 and 0.45, which is exactly the assumed separation, not a test of it. The paper gives no evidence that physical-layer trust mechanisms remain well-separated under an adaptive Byzantine adversary that can manipulate its transmitted signal. If the adversary can shift E[α_vu] - 0.5 to be nonnegative, the trusted set includes Byzantine clients and Eq. (17) no longer excludes their updates, breaking the regret bound. This needs either a formal argument for Assumption 1, a sensitivity analysis, or a clear statement tha
  4. [Section V, Figures 1 and 2] The simulation study is too narrow to substantiate the paper's claims. It considers exactly one graph topology (fully connected), one learning task (online logistic regression), one trust configuration (means 0.55/0.45, l=0.8), and one Byzantine fraction (30 of 45). No confidence intervals or error bars are shown, and the comparison to the Byzantine-free baseline is qualitative ('close') rather than quantitative. The paper should report numeric regret/violation values, standard deviations across the 50 realizations, and ideally vary the topology, the trust-separation gap, and the Byzantine fraction to show when the algorithm degrades.
minor comments (5)
  1. [Eq. (12)] The definition of β_vu(k) uses the index k both as the left-side argument and as the summation variable; the right-hand side depends on t, not k. Please rewrite as β_vu(t) = Σ_{k=0}^{t-1} (α_vu(k) - 0.5).
  2. [Lemma 1, Eq. (14)] The indicator term in (14) is written `I_{Em≥0}`; this appears to be a typo for `I_{Eb≥0}`. Similar notation issues occur around Ev and Eb in Eq. (11).
  3. [Eq. (24)] The logistic regression objective is missing a closing parenthesis: `log(1 + exp(-lv,t ψ^T_{v,t} x_{v,t}))`. Also κ_vu is used in the constraint but never defined.
  4. [Eqs. (5) and (6)] There is a typo in Eq. (5): the second term writes `fv.t` instead of `fv,t`. In Eq. (6), the notation `E/K` is not standard for edge-set difference; please use `E \ K` or define it explicitly.
  5. [Section V, paragraph on trust generation] The interval for the uniform distribution is written as `[E[α_vu(t)] - l/2, E[α_vu(t) + l/2]`; the second endpoint should be `E[α_vu(t)] + l/2`.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity; central guarantee is conditional on an explicitly stated trust-separation assumption and an external proof template.

full rationale

The paper's derivation chain is not circular. Assumption 1(ii) (E[alpha_vu] - 0.5 >= 0 for honest neighbors and < 0 for Byzantine neighbors) is explicitly introduced as an assumption from prior work [24], not as a consequence of the regret bound. Algorithm 1 filters neighbors using beta_vu(t) >= 0, and Lemma 1/Corollary 1 are standard concentration consequences of Assumption 1; they do not presuppose the theorem. The Theorem's O(sqrt(T)) regret and O(T^{3/4}) violation bounds are stated as a modification of the proof of Theorem 1 in [32], an external reference, so the central result is not reduced to the paper's own inputs. The simulation uses trust distributions with means 0.55/0.45, which instantiates Assumption 1; this is a limitation of the empirical validation (it does not validate physical-layer trust derivation) but it is not a fitted parameter renamed as a prediction. The only self-citation ([17]) supports a background statement in the introduction and is not load-bearing. The manuscript does contain an explicit omission: the proof of the Theorem is omitted due to space constraints and deferred to [32]'s proof. That is a verifiability/correctness gap, not circularity, since [32] is independent and the trust mechanism is not defined in terms of the target bounds. On the circularity axis, the paper receives a low score.

Assumptions & free parameters 3 free parameters · 8 assumptions · 0 invented entities

The paper's guarantees rest on six explicit assumptions, the most fragile being Assumption 1(ii) which asserts that trust scores separate honest from Byzantine neighbors in expectation. Without this separation, the trusted-neighbor filter (Step 9 of Algorithm 1) cannot distinguish adversaries, and the theorem's bound is moot. The other assumptions are standard convexity/smoothness conditions inherited from the cited online optimization framework.

free parameters (3)
  • Trust score generation parameters (E[alpha_honest]=0.55, E[alpha_byz]=0.45, l=0.8) = 0.55, 0.45, 0.8
    Set by the authors in the simulation (Section V) to create a separation of 0.05 from the 0.5 threshold; these are not estimated from data and directly determine the speed of misclassification decay.
  • Step-size constant a = implicitly 1 (since eta = 1/sqrt(T))
    In the theorem eta = a/sqrt(T) with a > 0; the simulation uses eta = 1/sqrt(T), so a = 1 is chosen without discussion.
  • Regularization delta = delta = 1/(4*eta^2)
    Chosen in the theorem to suppress dual variable growth; a design choice that affects the bounds.
assumptions (8)
  • domain assumption Assumption 1(i): The subgraph formed by honest clients is connected.
    Required so information can propagate among honest clients without passing through Byzantine nodes; stated in Section III-A.
  • domain assumption Assumption 1(ii): Trust scores satisfy E[alpha_vu] - 0.5 >= 0 for honest neighbors and < 0 for Byzantine neighbors.
    The core enabling condition for the trusted-neighbor set to eventually equal the honest set; stated in equation (11).
  • domain assumption Assumption 1(iii): Trust variables are independent and i.i.d.
    Used in Lemma 1's Hoeffding-style concentration bounds.
  • standard math Assumption 2: Feasible set X is convex, closed, bounded.
    Standard for online convex optimization; stated in Section IV.
  • standard math Assumption 3: Loss functions and constraints are convex.
    Needed for the saddle-point analysis; stated in Section IV.
  • standard math Assumption 4: Loss functions are G-Lipschitz.
    Bounded gradient magnitude assumption; equation (19).
  • standard math Assumption 5: Constraint gradients are L-Lipschitz.
    Smoothness of constraint functions; equations (20)-(21).
  • standard math Assumption 6: Existence of a point with bounded constraint values and gradients.
    Slater-type condition used in constraint violation bounds; stated in Section IV.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Online Decentralized Federated Multi-task Learning With Trustworthiness in Cyber-Physical Systems." pith.science (2026). https://pith.science/paper/ALCIUBUD

@misc{pith2026250900992,
  author       = {Pith},
  title        = {Pith review of: Online Decentralized Federated Multi-task Learning With Trustworthiness in Cyber-Physical Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ALCIUBUD}},
  note         = {Machine review of arXiv:2509.00992}
}
read the original abstract

Multi-task learning is an effective way to address the challenge of model personalization caused by high data heterogeneity in federated learning. However, extending multi-task learning to the online decentralized federated learning setting is yet to be explored. The online decentralized federated learning setting considers many real-world applications of federated learning, such as autonomous systems, where clients communicate peer-to-peer and the data distribution of each client is time-varying. A more serious problem in real-world applications of federated learning is the presence of Byzantine clients. Byzantine-resilient approaches used in federated learning work only when the number of Byzantine clients is less than one-half the total number of clients. Yet, it is difficult to put a limit on the number of Byzantine clients within a system in reality. However, recent work in robotics shows that it is possible to exploit cyber-physical properties of a system to predict clients' behavior and assign a trust probability to received signals. This can help to achieve resiliency in the presence of a dominating number of Byzantine clients. Therefore, in this paper, we develop an online decentralized federated multi-task learning algorithm to provide model personalization and resiliency when the number of Byzantine clients dominates the number of honest clients. Our proposed algorithm leverages cyber-physical properties, such as the received signal strength in wireless systems or side information, to assign a trust probability to local models received from neighbors in each iteration. Our simulation results show that the proposed algorithm performs close to a Byzantine-free setting.

Figures

Figures reproduced from arXiv: 2509.00992 by the authors.

Figure 1
Figure 1. Time-Average regret [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Time-Average constraint violation. by solving the following optimization problem among honest clients, min ⃗xv,t∈X ,∀v,t X T t=1 V X−b i=1  log(1 + exp(−lv,tψ T v,t⃗xv,t)  subject to 1 T X T t=1 ||⃗xv,t − ⃗xu,t||2 ≤ κ 2 vu; ∀(v, u) ∈ E h v (t) (24) where Dv,t := (lv,t, ψv,t) is the data sample of client v with label lv,t ∈ {1, −1} and feature vector ψv,t ∈ R d is drawn from the Gaussian distribution N (0, I). The … view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 29 canonical work pages

  1. [32]

    Decentralized multitask online convex optimization under random link failures,

    W. Yan and X. Cao, “Decentralized multitask online convex optimization under random link failures,” IEEE Transactions on Signal Processing , vol. 72, pp. 622–635, 2024

  2. [23]

    Resilience to malicious activity in distributed optimization for cyberphysical systems,

    M. Yemini, A. Nedi ´c, S. Gil, and A. J. Goldsmith, “Resilience to malicious activity in distributed optimization for cyberphysical systems,” in 2022 IEEE 61st Conference on Decision and Control (CDC) . IEEE, 2022, pp. 4185–4192

  3. [24]

    Resilient distributed optimization for multi-agent cyberphysical systems,

    M. Yemini, A. Nedi ´c, A. J. Goldsmith, and S. Gil, “Resilient distributed optimization for multi-agent cyberphysical systems,” IEEE Transactions on Automatic Control , 2025

  4. [1]

    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,” in Artificial intelligence and statistics . PMLR, 2017, pp. 1273– 1282

  5. [2]

    Federated learning on the road autonomous controller design for connected and autonomous vehicles,

    T. Zeng, O. Semiari, M. Chen, W. Saad, and M. Bennis, “Federated learning on the road autonomous controller design for connected and autonomous vehicles,” IEEE Transactions on Wireless Communications, vol. 21, no. 12, pp. 10 407–10 423, 2022

  6. [3]

    Federated learning in computer vision,

    D. Shenaj, G. Rizzoli, and P. Zanuttigh, “Federated learning in computer vision,” Ieee Access, vol. 11, pp. 94 863–94 884, 2023

  7. [4]

    Recovering private text in federated learning of language models,

    S. Gupta, Y . Huang, Z. Zhong, T. Gao, K. Li, and D. Chen, “Recovering private text in federated learning of language models,” Advances in neural information processing systems , vol. 35, pp. 8130–8143, 2022

  8. [5]

    Decentralized federated learning: Fundamentals, state of the art, frameworks, trends, and challenges,

    E. T. M. Beltr ´an, M. Q. P ´erez, P. M. S. S ´anchez, S. L. Bernal, G. Bovet, M. G. P ´erez, G. M. P ´erez, and A. H. Celdr ´an, “Decentralized federated learning: Fundamentals, state of the art, frameworks, trends, and challenges,” IEEE Communications Surveys & Tutorials , vol. 25, no. 4, pp. 2983–3013, 2023

Show all 32 references
  1. [6]

    Fully decentralized federated learning,

    A. Lalitha, S. Shekhar, T. Javidi, and F. Koushanfar, “Fully decentralized federated learning,” in Third workshop on bayesian deep learning (NeurIPS), vol. 12, 2018

  2. [7]

    Dynamic federated learning,

    E. Rizk, S. Vlaski, and A. H. Sayed, “Dynamic federated learning,” in 2020 IEEE 21st International Workshop on Signal Processing Advances in Wireless Communications (SPAWC). IEEE, 2020, pp. 1–5

  3. [8]

    Uav-enabled federated learning in dynamic environments: Efficiency and security trade-off,

    X. Fan, Y . Chen, M. Liu, S. Sun, Z. Liu, K. Xu, and Z. Li, “Uav-enabled federated learning in dynamic environments: Efficiency and security trade-off,” IEEE Transactions on Vehicular Technology , vol. 73, no. 5, pp. 6993–7006, 2023

  4. [9]

    Flexible training and uploading strategy for asynchronous federated learning in dynamic environments,

    M. Wu, M. Boban, and F. Dressler, “Flexible training and uploading strategy for asynchronous federated learning in dynamic environments,” IEEE Transactions on Mobile Computing , 2024

  5. [10]

    Federated learning with hierarchical clustering of local updates to improve training on non-iid data,

    C. Briggs, Z. Fan, and P. Andras, “Federated learning with hierarchical clustering of local updates to improve training on non-iid data,” in 2020 international joint conference on neural networks (IJCNN). IEEE, 2020, pp. 1–9

  6. [11]

    Personalized federated learning with theoretical guarantees: A model-agnostic meta-learning approach,

    A. Fallah, A. Mokhtari, and A. Ozdaglar, “Personalized federated learning with theoretical guarantees: A model-agnostic meta-learning approach,” Advances in neural information processing systems , vol. 33, pp. 3557–3568, 2020

  7. [12]

    Federated multi-task learning,

    V . Smith, C.-K. Chiang, M. Sanjabi, and A. S. Talwalkar, “Federated multi-task learning,” Advances in neural information processing systems, vol. 30, 2017

  8. [13]

    Federated multi-task learning under a mixture of distributions,

    O. Marfoq, G. Neglia, A. Bellet, L. Kameni, and R. Vidal, “Federated multi-task learning under a mixture of distributions,” Advances in Neural Information Processing Systems , vol. 34, pp. 15 434–15 447, 2021

  9. [14]

    Variational federated multi- task learning,

    L. Corinzia, A. Beuret, and J. M. Buhmann, “Variational federated multi- task learning,” arXiv preprint arXiv:1906.06268 , 2019

  10. [15]

    Challenges and approaches for mitigating byzantine attacks in federated learning,

    J. Shi, W. Wan, S. Hu, J. Lu, and L. Y . Zhang, “Challenges and approaches for mitigating byzantine attacks in federated learning,” in 2022 IEEE International Conference on Trust, Security and Privacy in Computing and Communications (TrustCom) . IEEE, 2022, pp. 139– 146

  11. [16]

    Local model poisoning attacks to {Byzantine-Robust} federated learning,

    M. Fang, X. Cao, J. Jia, and N. Gong, “Local model poisoning attacks to {Byzantine-Robust} federated learning,” in 29th USENIX security symposium (USENIX Security 20) , 2020, pp. 1605–1622

  12. [17]

    Online decentralized multi- agents meta-learning with byzantine resiliency,

    O. T. Odeyomi, B. Ude, and K. Roy, “Online decentralized multi- agents meta-learning with byzantine resiliency,” IEEE Access , vol. 11, pp. 68 286–68 300, 2023

  13. [18]

    Geometric median in nearly linear time,

    M. B. Cohen, Y . T. Lee, G. Miller, J. Pachocki, and A. Sidford, “Geometric median in nearly linear time,” in Proceedings of the forty- eighth annual ACM symposium on Theory of Computing, 2016, pp. 9–21

  14. [19]

    Byzantine-robust dis- tributed learning: Towards optimal statistical rates,

    D. Yin, Y . Chen, R. Kannan, and P. Bartlett, “Byzantine-robust dis- tributed learning: Towards optimal statistical rates,” in International conference on machine learning . Pmlr, 2018, pp. 5650–5659

  15. [20]

    Bridge: Byzantine-resilient decen- tralized gradient descent,

    C. Fang, Z. Yang, and W. U. Bajwa, “Bridge: Byzantine-resilient decen- tralized gradient descent,” IEEE Transactions on Signal and Information Processing over Networks , vol. 8, pp. 610–626, 2022

  16. [21]

    Byzantine-resilient decentralized stochastic gradient descent,

    S. Guo, T. Zhang, H. Yu, X. Xie, L. Ma, T. Xiang, and Y . Liu, “Byzantine-resilient decentralized stochastic gradient descent,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 32, no. 6, pp. 4096–4106, 2021

  17. [22]

    Byrdie: Byzantine-resilient distributed coordinate descent for decentralized learning,

    Z. Yang and W. U. Bajwa, “Byrdie: Byzantine-resilient distributed coordinate descent for decentralized learning,” IEEE Transactions on Signal and Information Processing over Networks , vol. 5, no. 4, pp. 611–627, 2019

  18. [25]

    Guaranteeing spoof-resilient multi-robot networks,

    S. Gil, S. Kumar, M. Mazumder, D. Katabi, and D. Rus, “Guaranteeing spoof-resilient multi-robot networks,” Autonomous Robots, vol. 41, pp. 1383–1400, 2017

  19. [26]

    Cp-mtml: Coupled projection multi-task metric learning for large scale face retrieval,

    B. Bhattarai, G. Sharma, and F. Jurie, “Cp-mtml: Coupled projection multi-task metric learning for large scale face retrieval,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2016, pp. 4226–4235

  20. [27]

    Proximity without consen- sus in online multiagent optimization,

    A. Koppel, B. M. Sadler, and A. Ribeiro, “Proximity without consen- sus in online multiagent optimization,” IEEE Transactions on Signal Processing, vol. 65, no. 12, pp. 3062–3077, 2017

  21. [28]

    Asynchronous online learning in multi-agent systems with proximity constraints,

    A. S. Bedi, A. Koppel, and K. Rajawat, “Asynchronous online learning in multi-agent systems with proximity constraints,” IEEE Transactions on Signal and Information Processing over Networks , vol. 5, no. 3, pp. 479–494, 2019

  22. [29]

    Online learning with sample path constraints

    S. Mannor, J. N. Tsitsiklis, and J. Y . Yu, “Online learning with sample path constraints.” Journal of Machine Learning Research, vol. 10, no. 3, 2009

  23. [30]

    Safety-aware algorithms for adversarial contextual bandit,

    W. Sun, D. Dey, and A. Kapoor, “Safety-aware algorithms for adversarial contextual bandit,” in International Conference on Machine Learning . PMLR, 2017, pp. 3280–3288

  24. [31]

    Characterizing trust and resilience in distributed consensus for cyberphysical systems,

    M. Yemini, A. Nedi ´c, A. J. Goldsmith, and S. Gil, “Characterizing trust and resilience in distributed consensus for cyberphysical systems,” IEEE Transactions on Robotics , vol. 38, no. 1, pp. 71–91, 2021

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.