pith. sign in

arxiv: 2606.28441 · v1 · pith:XFLLZYEBnew · submitted 2026-06-26 · 💻 cs.LG · cs.AI· cs.MA· cs.SY· eess.SY

Learning to Distributedly Estimate under Partially Known Dynamics: A Covariance-Agnostic Neural Kalman Consensus Filter

George Stamatelis , Kyriakos Stylianopoulos , George C. Alexandropoulos This is my paper

Pith reviewed 2026-06-30 01:23 UTC · model grok-4.3

classification 💻 cs.LG cs.AIcs.MAcs.SYeess.SY
keywords distributed state estimationneural Kalman filterconsensus filtercovariance agnosticmulti-agent systemslatent state estimationmodel misspecification
0
0 comments X

The pith

Agents collaboratively estimate latent states by learning consensus weights and Kalman-like updates without any noise statistics knowledge.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces a distributed sensing framework called CA-NKCF where multiple agents exchange information to estimate hidden system states. It integrates partial knowledge of the motion and observation models with deep neural networks that learn to set consensus weights and perform recursive updates. Experiments on linear systems, Lorenz chaos, and wireless tracking show the method exceeds the accuracy of both classical distributed Kalman and particle filters and purely data-driven networks, even when the supplied models are inaccurate.

Core claim

The Covariance-Agnostic Neural Kalman Consensus Filter performs decentralized latent state estimation by feeding prior estimates into neural networks that compute optimized consensus weights and Kalman-style correction steps, all without access to noise covariance matrices; the resulting estimator remains accurate under model misspecification across linear, chaotic, and practical wireless scenarios.

What carries the argument

The CA-NKCF, a neural-network-driven variant of the Kalman consensus filter that learns consensus weights and update gains from partial dynamics knowledge to replace explicit covariance calculations.

If this is right

  • The estimator maintains its advantage over baselines across varying noise intensities, random communication graphs, state dimensions, and clutter densities.
  • Accuracy holds when the motion and observation models supplied to the agents are misspecified.
  • The same learned structure applies to both linear and nonlinear dynamics such as the Lorenz attractor.
  • The approach supports online tasks including change-point detection without requiring separate noise calibration.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The method could be tested on physical multi-robot platforms where sensor noise is hard to characterize in advance.
  • Training the networks on data from one topology and deploying on another would check generalization of the learned consensus rules.
  • Extending the framework to include intermittent communication or packet loss would reveal whether the neural updates remain stable under realistic network faults.

Load-bearing premise

Neural networks can be trained to produce reliable consensus weights and update steps from partial dynamics knowledge alone when noise statistics are entirely unavailable.

What would settle it

A controlled simulation in which the neural filter's mean squared error exceeds that of a standard distributed particle filter once the state dimension exceeds 10 and the supplied motion model deviates by more than 30 percent from truth.

Figures

Figures reproduced from arXiv: 2606.28441 by George C. Alexandropoulos, George Stamatelis, Kyriakos Stylianopoulos.

Figure 1
Figure 1. Figure 1: Visualization of the proposed CA-NKCF framework for distributed estimation with [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Sampled oscillator states for N = 4 sensors. optimization objective for the considered NN-based distributed estimation framework: min θ,γ L ≜ 1 N X N i=1 Li . (19) The proposed model was trained on the mini-batch version of L. For each trajectory in the batch, the neural filtering operation from t = 0 up to t = T was conducted sequentially, and the local losses at each time instance t were added to compute… view at source ↗
Figure 3
Figure 3. Figure 3: MSE performance versus the node connection probability for the linear scenario. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Average MSE versus the number of nodes N for the linear scenario. 2) One important feature of our method is the joint opti￾mization of γ. To investigate whether joint optimization hinders stability, we have trained the same models, but with a fixed γ, whose value was determined by an element-wise grid search on [0, 1]. The training process was repeated for each considered value of γ. 3) For the final ablat… view at source ↗
Figure 5
Figure 5. Figure 5: The considered state trajectory for the Lorenz attractor [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Average MSE versus the observation noise level for [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 8
Figure 8. Figure 8: Average MSE versus the number of scatterers for the [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Sensitivity study results for the wireless UE tracking [PITH_FULL_IMAGE:figures/full_fig_p010_9.png] view at source ↗
read the original abstract

Online latent state estimation constitutes a fundamental challenge within the artificial intelligence field, serving as a foundational tool for diverse applications, including sequential decision making, anomaly and change-point detection. In this paper, a novel online distributed sensing framework, where agents collaborate and exchange information to perform latent state estimation, is presented. The proposed estimator combines available partial domain knowledge with the representation capabilities of deep neural networks. In particular, the designed sensing framework incorporates prior estimates, optimized consensus weights, and Kalman-like recursive updates to perform decentralized inference, without relying on knowledge of noise statistics. Extensive experiments on linear, chaotic (Lorenz), and practical wireless tracking environments reveal that the proposed Covariance-Agnostic Neural Kalman Consensus Filter (CA-NKCF) outperforms traditional distributed Kalman and particle filters as well as purely model-free deep neural networks, exhibiting robustness even when the underlying motion and observation models are misspecified. It is also demonstrated that CA-NKCF's performance advantage remains stable across varying noise levels, random communication topologies, latent state dimensions, and observation clutter densities induced by scattering objects in wireless systems.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 3 minor

Summary. The paper proposes the Covariance-Agnostic Neural Kalman Consensus Filter (CA-NKCF), a hybrid distributed estimation framework that fuses partial domain knowledge of dynamics with deep neural networks. The method incorporates prior estimates, learned consensus weights, and Kalman-style recursive updates to perform decentralized latent state inference without access to noise covariance statistics. Experiments across linear systems, Lorenz chaotic dynamics, and wireless tracking scenarios report consistent outperformance relative to distributed Kalman filters, particle filters, and purely model-free neural networks, with maintained advantages under model misspecification, varying noise levels, random topologies, state dimensions, and clutter densities.

Significance. If the empirical results hold, the work demonstrates a practical route to covariance-agnostic distributed estimation by combining partial physics with learned components, addressing a common limitation in sensor networks and tracking applications. Credit is due for the breadth of experimental validation (linear, nonlinear chaotic, and wireless cases) and explicit robustness tests under misspecification; these provide concrete evidence beyond abstract claims.

minor comments (3)
  1. [§3] §3 (method): the precise architecture of the neural modules that output consensus weights and correction gains should be stated explicitly (layer counts, activation functions, input features) so that the hybrid construction can be reproduced without ambiguity.
  2. [§4] §4 (experiments): while outperformance is reported, the tables or figures should include standard deviations or confidence intervals across random seeds and communication graphs to substantiate the stability claims across topologies.
  3. Notation: the distinction between the learned quantities and the partial model-based terms (e.g., the prediction step) should be clarified with a single summary equation or table to avoid reader confusion between the two information sources.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive assessment of our manuscript on the Covariance-Agnostic Neural Kalman Consensus Filter (CA-NKCF) and for recommending minor revision. The summary accurately reflects the hybrid neural-Kalman approach, its covariance-agnostic property, and the breadth of experimental validation across linear, chaotic, and wireless settings.

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The manuscript presents a hybrid neural architecture for distributed state estimation that fuses partial dynamics knowledge with learned consensus weights and recursive updates, remaining agnostic to noise covariances. All load-bearing claims rest on the training procedure and empirical outperformance across linear, Lorenz, and wireless scenarios under misspecification; no equations, uniqueness theorems, or self-citations are invoked to derive performance guarantees by construction. The reported results are therefore falsifiable against external benchmarks and do not reduce to fitted inputs renamed as predictions or to self-referential definitions.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides no equations, no parameter lists, and no explicit assumptions beyond the high-level description of the framework; cannot populate free parameters, axioms, or invented entities.

pith-pipeline@v0.9.1-grok · 5745 in / 1140 out tokens · 32949 ms · 2026-06-30T01:23:54.050550+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

65 extracted references · 6 canonical work pages · 6 internal anchors

  1. [1]

    LumiMAS: A comprehensive framework for real- time monitoring and enhanced observability in multi-agent systems,

    R. Solomonet al., “LumiMAS: A comprehensive framework for real- time monitoring and enhanced observability in multi-agent systems,” in Proc. AAMAS, (Paphos, Cyprus), 2026

    2026

  2. [2]

    IMAS2: Joint agent selection and information-theoretic coordinated perception in Dec-POMDPs,

    C. Shiet al., “IMAS2: Joint agent selection and information-theoretic coordinated perception in Dec-POMDPs,” inProc. AAMAS, (Paphos, Cyprus), 2026

    2026

  3. [3]

    An approach to target tracking,

    M. Gruber, “An approach to target tracking,” technical report, MIT Lincoln Laboratory, Lexington, MA, 1967

    1967

  4. [4]

    Application of the extended Kalman filter to ballistic trajectory estimation,

    R. E. Larsonet al., “Application of the extended Kalman filter to ballistic trajectory estimation,”Stanford Research Institute, Tech. Rep., 1967

    1967

  5. [5]

    Industrial applications of the Kalman filter: A review,

    F. Augeret al., “Industrial applications of the Kalman filter: A review,” IEEE Trans. Ind. Electron, vol. 60, no. 12, pp. 5458–5471, 2013

    2013

  6. [6]

    A distributed control framework for a team of unmanned aerial vehicles for dynamic wildfire tracking,

    H. X. Pham and Lothers, “A distributed control framework for a team of unmanned aerial vehicles for dynamic wildfire tracking,” inProc. IEEE/RSJ IROS, (Vancouver, Canada), 2017

    2017

  7. [7]

    Simultaneous distributed acoustic and temperature sensing for robust leakage detection in gas pipelines,

    V . P. Anandet al., “Simultaneous distributed acoustic and temperature sensing for robust leakage detection in gas pipelines,”J. Lightwave Techn., vol. 44, no. 7, pp. 2849–2857, 2026

    2026

  8. [8]

    A new approach to linear filtering and prediction problems,

    R. E. Kalman, “A new approach to linear filtering and prediction problems,”J. Basic Engineering, vol. 82, pp. 35–45, 03 1960

    1960

  9. [9]

    The unscented Kalman filter for nonlinear estimation,

    E. Wan and R. Van Der Merwe, “The unscented Kalman filter for nonlinear estimation,” inProc. IEEE AS-SPCC, (Alberta, Canada), 2000

    2000

  10. [10]

    CTD4 – a deep continuous distributional actor- critic agent with a Kalman fusion of multiple critics,

    D. Valenciaet al., “CTD4 – a deep continuous distributional actor- critic agent with a Kalman fusion of multiple critics,” inProc. AAAI, (Philadelphia, Pennsylvania, USA), 2025

    2025

  11. [11]

    Partial diffusion Kalman filtering for distributed state estimation in multiagent networks,

    V . Vahidpouret al., “Partial diffusion Kalman filtering for distributed state estimation in multiagent networks,”IEEE Trans. Neural Netw. Learn. Syst., vol. 30, no. 12, pp. 3839–3846, 2019

    2019

  12. [12]

    Artificial intelligence-aided Kalman filters: AI- augmented designs for Kalman-type algorithms,

    N. Shlezingeret al., “Artificial intelligence-aided Kalman filters: AI- augmented designs for Kalman-type algorithms,”IEEE Signal Process. Mag., pp. 2–26, 2025

    2025

  13. [13]

    Long short-term memory,

    S. Hochreiter and J. Schmidhuber, “Long short-term memory,”Neural Computation, vol. 9, no. 8, pp. 1735–1780, 1997

    1997

  14. [14]

    Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling

    J. Chunget al., “Empirical evaluation of gated recurrent neural networks on sequence modeling,”arXiv preprint: 1412.3555, 2014

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  15. [15]

    Identification of nonlinear state space models using an MLP network trained by the EM algorithm,

    A. A. Gorji and M. B. Menhaj, “Identification of nonlinear state space models using an MLP network trained by the EM algorithm,” inProc. IEEE IJCNN, (Hong Kong, China), 2008

    2008

  16. [16]

    Sequential neural models with stochastic layers,

    M. Fraccaroet al., “Sequential neural models with stochastic layers,” in Proc. NeurIPS, (Barcelona, Spain), 2016

    2016

  17. [17]

    Combining generative and discriminative models for hybrid inference,

    V . G. Satorraset al., “Combining generative and discriminative models for hybrid inference,” inProc. NeurIPS, (Vancouver Canada), 2019

    2019

  18. [18]

    DANSE: Data-driven non-linear state estimation of model-free process in unsupervised learning setup,

    A. Ghoshet al., “DANSE: Data-driven non-linear state estimation of model-free process in unsupervised learning setup,”IEEE Trans. Signal Process., vol. 72, pp. 1824–1838, 2024

    2024

  19. [19]

    Long short-term memory Kalman filters: Recurrent neural estimators for pose regularization,

    H. Coskunet al., “Long short-term memory Kalman filters: Recurrent neural estimators for pose regularization,” inProc. IEEE ICCV, (Venice, Italy), 2017

    2017

  20. [20]

    KalmanNet: Neural network aided Kalman filtering for partially known dynamics,

    G. Revachet al., “KalmanNet: Neural network aided Kalman filtering for partially known dynamics,”IEEE Trans. Signal Proces., vol. 70, pp. 1532–1547, 2022

    2022

  21. [21]

    Latent-KalmanNet: Learned Kalman filtering for tracking from high-dimensional signals,

    I. Buchniket al., “Latent-KalmanNet: Learned Kalman filtering for tracking from high-dimensional signals,”IEEE Trans. Signal Process., vol. 72, pp. 352–367, 2024

    2024

  22. [22]

    Filtering Markov jump systems with partially known dynamics: A model-based deep learning approach,

    G. Stamatelis and G. C. Alexandropoulos, “Filtering Markov jump systems with partially known dynamics: A model-based deep learning approach,”IEEE Trans. Signal Proces. (Early Acess), 2026

    2026

  23. [23]

    Distributed weighted average consensus fusion based on admm under measurement uncertainty,

    T. Cuiet al., “Distributed weighted average consensus fusion based on admm under measurement uncertainty,”Signal Process., vol. 241, p. 110380, 2026

    2026

  24. [24]

    Consensus in multi-agent systems: a review,

    A. Amirkhani and A. H. Barshooi, “Consensus in multi-agent systems: a review,”Artif. Intell. Rev., vol. 55, p. 3897–3935, June 2022

    2022

  25. [25]

    Kalman-consensus filter : Optimality, stability, and performance,

    R. Olfati-Saber, “Kalman-consensus filter : Optimality, stability, and performance,” inProc. IEEE CDC, (Shanghai, China), 2009

    2009

  26. [26]

    Multi-agent reinforcement learning as a rehearsal for decentralized planning,

    L. Kraemer and B. Banerjee, “Multi-agent reinforcement learning as a rehearsal for decentralized planning,”Neurocomputing, vol. 190, pp. 82– 94, 2016

    2016

  27. [27]

    Multi-agent actor-critic for mixed cooperative- competitive environments,

    R. Loweet al., “Multi-agent actor-critic for mixed cooperative- competitive environments,” inProc. NeurIPS, (Long Beach, California, USA), 2017

    2017

  28. [28]

    Distributed kalman filtering for sensor networks,

    R. Olfati-Saber, “Distributed kalman filtering for sensor networks,” in Proc. IEEE CDC, (New Orleans, Louisiana, USA), 2007

    2007

  29. [29]

    Information weighted consensus,

    A. T. Kamalet al., “Information weighted consensus,” inProc. IEEE CDC, (Maui, Hawaii, USA), 2012

    2012

  30. [30]

    A fully decentralized multi-sensor system for tracking and surveillance,

    B. Raoet al., “A fully decentralized multi-sensor system for tracking and surveillance,”Intern. J. Robotics Research, vol. 12, no. 1, pp. 20–44, 1993

    1993

  31. [31]

    Optimal discrete-time kalman consensus filter,

    R. Deshmukhet al., “Optimal discrete-time kalman consensus filter,” in Proc. ACC, (Seattle, Washington, USA), 2017

    2017

  32. [32]

    Optimal Kalman filter with information-weighted consensus,

    S. Khanet al., “Optimal Kalman filter with information-weighted consensus,”IEEE Trans. Autom. Control, vol. 68, no. 9, pp. 5624–5629, 2023

    2023

  33. [33]

    An efficient distributed Kalman filter over sensor networks with maximum correntropy criterion,

    C. Hu and B. Chen, “An efficient distributed Kalman filter over sensor networks with maximum correntropy criterion,”IEEE Trans. Signal Inf. Process. Networks, vol. 8, pp. 433–444, 2022

    2022

  34. [34]

    Distributed Kalman consensus filter for estimation with moving targets,

    B. Lianet al., “Distributed Kalman consensus filter for estimation with moving targets,”IEEE Trans. Cybern., vol. 52, no. 6, pp. 5242–5254, 2022

    2022

  35. [35]

    Distributed Kalman filtering under two-bitrate peri- odic coding strategies,

    Q. Liuet al., “Distributed Kalman filtering under two-bitrate peri- odic coding strategies,”IEEE Trans. Autom. Control, vol. 69, no. 12, pp. 8633–8646, 2024

    2024

  36. [36]

    Distributed Kalman filtering with event- triggered communication: a robust approach,

    D. Ghion and M. Zorzi, “Distributed Kalman filtering with event- triggered communication: a robust approach,” inProc. Medit. Conf. Control Autom., (Athens, Greece), 2022

    2022

  37. [37]

    Privacy-preserving distributed Kalman filtering,

    A. Moradiet al., “Privacy-preserving distributed Kalman filtering,”IEEE Trans. Signal Process., vol. 70, pp. 3074–3089, 2022

    2022

  38. [38]

    Particle learning and smoothing,

    C. M. Carvalhoet al., “Particle learning and smoothing,”Statistical Science, vol. 25, no. 1, pp. 88–106, 2010

    2010

  39. [39]

    Sequential monte carlo methods under model uncer- tainty,

    I. Urtcagaet al., “Sequential monte carlo methods under model uncer- tainty,” inProc. IEEE SSP, (Palma de Mallorca, Spain), 2016

    2016

  40. [40]

    A particle filtering based approach to approximating interactive POMDPs,

    P. Doshi and P. J. Gmytrasiewicz, “A particle filtering based approach to approximating interactive POMDPs,” inProc. AAAI, (Pittsburgh, Pennsylvania, USA), 2005

    2005

  41. [41]

    Distributed particle filters for sensor networks,

    M. Coates, “Distributed particle filters for sensor networks,” inProc. ACM IPSN, (Berkeley, California, USA), 2004

    2004

  42. [42]

    RTSNet: Learning to smooth in partially known state- space models,

    G. Revachet al., “RTSNet: Learning to smooth in partially known state- space models,”IEEE Trans. Signal Process., vol. 71, pp. 4441–4456, 2023

    2023

  43. [43]

    Structured inference networks for nonlinear state space models,

    R. G. Krishnanet al., “Structured inference networks for nonlinear state space models,” inProc. AAAI, (San Francisco, California, USA), 2017

    2017

  44. [44]

    Deep Kalman Filters

    R. G. Krishnanet al., “Deep Kalman filters,”arXiv preprint: 1511.05121, 2015

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  45. [45]

    A disentangled recognition and nonlinear dynamics model for unsupervised learning,

    M. Fraccaroet al., “A disentangled recognition and nonlinear dynamics model for unsupervised learning,” inProc. NeurIPS, (Long Beach, California, USA), 2017

    2017

  46. [46]

    Model-based deep learning,

    N. Shlezingeret al., “Model-based deep learning,”Proc. IEEE, vol. 111, no. 5, pp. 465–499, 2023

    2023

  47. [47]

    MoDL: Model-based deep learning architecture for inverse problems,

    H. K. Aggarwalet al., “MoDL: Model-based deep learning architecture for inverse problems,”IEEE Trans. Medical Imag., vol. 38, no. 2, pp. 394–405, 2019

    2019

  48. [48]

    DCD-MUSIC: Deep-learning-aided cascaded differen- tiable MUSIC algorithm for near-field localization of multiple sources,

    A. Gastet al., “DCD-MUSIC: Deep-learning-aided cascaded differen- tiable MUSIC algorithm for near-field localization of multiple sources,” inProc. IEEE ICASSP, (Hyderabad, India), 2025

    2025

  49. [49]

    Learned robust PCA: a scalable deep unfolding approach for high-dimensional outlier detection,

    H. Caiet al., “Learned robust PCA: a scalable deep unfolding approach for high-dimensional outlier detection,” inProc. NeurIPS, (Virtual), 2021

    2021

  50. [50]

    Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,

    M. Raissiet al., “Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,”J. Comput. Physics, vol. 378, pp. 686–707, 2019

    2019

  51. [51]

    Dual-balancing for physics-informed neural networks,

    C. Zhouet al., “Dual-balancing for physics-informed neural networks,” inProc. IJCAI, (Montreal Canada), 2025. 13

    2025

  52. [52]

    Graph-CNNs for RF imaging: Learning the electric field integral equations,

    K. Stylianopouloset al., “Graph-CNNs for RF imaging: Learning the electric field integral equations,” inProc. EUSIPCO, (Palermo, Italy), 2025

    2025

  53. [53]

    A Physics-Informed Hierarchical Neural Network for Microwave Scattering Analysis of 3D PEC Targets

    R. Zhuet al., “U-PINet: Physics-informed hierarchical learning for radar cross section prediction via 3D electromagnetic scattering reconstruc- tion,”arXiv preprint: 2508.03774, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  54. [54]

    Attention Is All You Need

    A. Vaswaniet al., “Attention is all you need,”arXiv preprint arXiv:1706.03762, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  55. [55]

    Dota 2 with Large Scale Deep Reinforcement Learning

    OpenAI, “Dota 2 with large scale deep reinforcement learning,”arXiv preprint: 1912.06680, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1912

  56. [56]

    Parameter sharing with network pruning for scalable multi-agent deep reinforcement learning,

    W. Kim and Y . Sung, “Parameter sharing with network pruning for scalable multi-agent deep reinforcement learning,” inProc. AAMAS, (London, United Kingdom), 2023

    2023

  57. [57]

    Parameter sharing reinforcement learning archi- tecture for multi agent driving,

    M. Kaushiket al., “Parameter sharing reinforcement learning archi- tecture for multi agent driving,” inProc. AIR, ACM International Conference Proceedings Series, (Chennai, India), 2019

    2019

  58. [58]

    Consensus seeking in multiagent systems under dynamically changing interaction topologies,

    W. Ren and R. Beard, “Consensus seeking in multiagent systems under dynamically changing interaction topologies,”IEEE Trans. Autom. Control, vol. 50, no. 5, pp. 655–661, 2005

    2005

  59. [59]

    Consensus and cooperation in networked multi- agent systems,

    R. Olfati-Saberet al., “Consensus and cooperation in networked multi- agent systems,”Proc. IEEE, vol. 95, no. 1, pp. 215–233, 2007

    2007

  60. [60]

    Adam: A Method for Stochastic Optimization

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint: 1412.6980, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  61. [61]

    Lstm: A search space odyssey,

    K. Greffet al., “Lstm: A search space odyssey,”IEEE Trans Neural Netw. Learn. Syst., vol. 28, no. 10, pp. 2222–2232, 2017

    2017

  62. [62]

    NR; Physical channels and modulation,

    3GPP, “NR; Physical channels and modulation,” Technical Specification TS 38.211, 3rd Generation Partnership Project (3GPP), 2020

    2020

  63. [63]

    Radio localization and sens- ing—Part i: Fundamentals,

    H. Wymeersch and G. Seco-Granados, “Radio localization and sens- ing—Part i: Fundamentals,”IEEE Commun. Lett., vol. 26, no. 12, pp. 2816–2820, 2022

    2022

  64. [64]

    A tutorial on 5G positioning,

    L. Italianoet al., “A tutorial on 5G positioning,”IEEE Commun. Surveys & Tuts., vol. 27, no. 3, pp. 1488–1535, 2025

    2025

  65. [65]

    Deconstructing multiantenna fading channels,

    A. M. Sayeed, “Deconstructing multiantenna fading channels,”IEEE Trans. Signal Process., vol. 50, no. 10, pp. 2563–2579, 2002

    2002