Pith. sign in

REVIEW 5 major objections 6 minor 1 cited by

Change-Aware Self-Adaptive AI-Aided Kalman Filters With Neural Change Point Detection

T0 review · 5 major / 6 minor · reviewed 2026-08-02 · deepseek-v4-flash

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

Pith's one-line read A learned reliability monitor on a Kalman filter's prediction residual can gate label-free online adaptation, letting AI-aided filters track abrupt and gradual model changes while approaching full-information accuracy.

desk verdict A sensible online-adaptation wrapper for KalmanNet that mostly delivers, but the evaluation omits the closest competitors and leaves the innovation-proxy assumption untested. read the letter →

arxiv 2607.13387 v1 pith:4Y7J2532 submitted 2026-07-15 eess.SY cs.SY

Wenyi Zhang , Xiaoyong Ni , Nir Shlezinger , Zengfu Wang This is my paper · ORCID
classification eess.SYcs.SY
keywords KalmanfilterstateestimationonlineadaptationchangepointdetectionmodelmismatchunsupervisedlearningNetnonstationarysystems
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 aims to make AI-aided Kalman filters — specifically KalmanNet, a neural version of the extended Kalman filter — self-adapt when the system dynamics or noise statistics drift. It argues that the filter's own prediction residual, viewed through a learned change-detection module, is enough to tell when the filter has lost alignment and how hard to update it. The resulting framework, CASA-KalmanNet, performs label-free online updates only when triggered, scaling update size by detected degradation. If this is right, it narrows the gap to optimal model-based filters in nonstationary settings while avoiding the over-adaptation that continuous unsupervised updating suffers from.

What carries the argument

CPDNet is the central mechanism: a recurrent network that consumes windowed scalar features z_t = tanh(γ(σ(‖Δy_t‖) − 0.5)) derived from the innovation norm. Its output p̂_φ feeds a gated learning rate η_t = ε·max(p̂_φ − thresh, 0), which switches on a single gradient step on the unsupervised, ℓ2-regularized loss L_t^unsup(θ) = ‖Δy_t(θ)‖² + ρ‖θ‖². Together these pieces convert an internal filter signal into a closed-loop, label-free adaptation rule that decides when to update, how strongly to update, and when to leave the pretrained filter alone.

What would settle it

Construct a model shift that reduces the observation gain (so the innovation magnitude shrinks even though state uncertainty grows) and check whether CPDNet stays silent while MSE degrades; alternatively, compare CASA-KalmanNet against an oracle adaptation triggered by true state error — if the oracle is measurably better across Q/R shifts, the innovation-proxy assumption is falsified.

Watch

Extended reading notes

Core claim

The central claim is that the interpretable internal features of KalmanNet — especially the innovation process Δy_t, the gap between observed measurement and the filter's predicted measurement — carry enough information to drive both detection of and recovery from distribution shift. CPDNet, a lightweight recurrent network, maps windowed normalized innovation magnitudes to a soft reliability-degradation score; that score is converted into a time-varying learning rate that gates and scales a single truncated-backpropagation gradient step on an unsupervised innovation-loss objective. This loop lets CASA-KalmanNet adapt to abrupt and gradual changes in noise statistics and model parameters with

Load-bearing premise

The load-bearing premise is that the size of the filter's prediction residual (the gap between observed and predicted measurements) tracks how badly the filter is failing, and that shrinking that residual online improves state estimation; if a regime change makes the residual stay small while state errors grow, the whole trigger-and-update loop chases the wrong signal.

Editorial extensions

If this is right

  • AI-aided KalmanNet-style filters can keep operating through abrupt and gradual noise or model changes without any ground-truth state labels from the new regime.
  • Because adaptation is gated and scaled by the reliability score, update counts drop sharply — 51.8% fewer updates than continuous unsupervised adaptation in the real-robot experiment — while accuracy still improves.
  • Under full parameter knowledge, CASA-KalmanNet approaches the MSE of the optimal Kalman and EM-Kalman filters on linear systems and beats the EKF, UKF, KalmanNet, and particle filter baselines on the nonlinear Lorenz system under model mismatch.
  • The mechanism extends beyond noise-covariance shifts to rotations of the state-transition and observation matrices, although the paper notes that major structural changes would require an additional system-identification layer.
  • On a controlled robot-trajectory test with a 500× unannounced jump in observation noise, CASA-KalmanNet achieves the best MSE among learned filters while making far fewer online updates.

Reading between the lines

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

  • The same gating idea could transfer to other neural filters only if they expose a comparable internal residual; the paper explicitly leaves that redesign open, so a natural extension is testing CPDNet-style monitoring on non-Kalman recursive estimators.
  • The paper's own stated limits suggest stress tests rather than settled results: very slow drift can delay or miss the CPDNet trigger, and repeated changes over long horizons may accumulate update error — both are worth probing empirically.
  • Because the unsupervised loss is the innovation norm, the method is most natural where observations remain informative; in extremely low-SNR regimes the update signal may become noise-dominated, a boundary the paper's −10 dB experiments only begin to explore.
  • A direct extension the paper lists as future work — adapting the detection window δ and the number of gradient steps to the detected severity — could reduce latency in fast-changing environments and deserves a dedicated study.
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

5 major / 6 minor

Summary. The paper proposes CASA-KalmanNet, an online test-time adaptation framework for KalmanNet. It adds CPDNet, a neural module that takes a window of nonlinear scalar features of the innovation norm (Eq. 10) and outputs a reliability-degradation score. This score gates and scales the learning rate (Eq. 15) for an online update of KalmanNet parameters using an unsupervised innovation-norm loss (Eq. 16). CPDNet is trained offline against the reference KalmanNet's state error (Eq. 13). Experiments cover linear Gaussian models with abrupt and gradual changes in Q, R, F, H; a Lorenz attractor nonlinear model; and an NCLT robot trajectory, comparing against KF/EM-KF/EKF/UKF/PF and supervised/unsupervised KalmanNet. The paper claims consistent improvement over learning-based baselines and near-optimal performance relative to full-information classical filters.

Significance. If the claims hold, the framework is a valuable step toward label-free online adaptation of AI-aided Kalman filters: it monitors filter reliability rather than the raw data distribution, updates only when needed, and reduces update count substantially (Table VII). Strengths include a clear architecture, a reproducible code repository, a real-data experiment, and honest statements of limitations in Sections III-B1, IV-C, and V-C. However, the central proxy assumption—that the innovation norm is both a valid trigger and a valid adaptation objective—is not empirically isolated, and the experimental protocol lacks error bars and the most relevant adaptive baselines. These gaps leave the principal claims stronger than the evidence.

major comments (5)
  1. [Section IV-A2, Eq. (16)] The unsupervised loss L_t^unsup(θ)=||Δy_t(θ)||^2+ρ||θ||^2 is the sole adaptation signal, but the paper never establishes that minimizing the innovation norm improves state estimation. In the linear Gaussian model, Δy_t = H(x_t - x̂_{t|t-1}) + v_t, so under large observation noise R a large innovation is dominated by v_t, and minimizing ||Δy_t||^2 can pull the prior estimate toward the noisy observation, increasing corrected state error. Section IV-A1 explicitly labels η_t a 'heuristic adaptation magnitude,' but no experiment isolates the proxy assumption. Please add an ablation that varies R (e.g., compares Eq. (16) against a normalized or residual-weighted loss) or otherwise demonstrates that the update direction is not harmful in noise-dominated regimes; otherwise the central adaptation mechanism is unvalidated.
  2. [Section III-C2/Eq. (13) and Section IV-A2/Eq. (17)] CPDNet is trained offline against the error of the fixed reference KalmanNet θ (Eq. 13), but at deployment it gates updates for the continually adapted network. After parameters change, the mapping from innovations to actual state error is no longer the one CPDNet learned; the paper does not recalibrate or re-evaluate CPDNet on the adapted network. This could lead to miscalibrated gating after the first update. Please report CPDNet's detection accuracy against the adapted network's true error over time, or compare against an oracle-gated version, to show the gating remains valid.
  3. [Section V-B, Tables II and III] The linear-model tables report only point MSE values, with no error bars, number of seeds, or significance tests. Since change points are randomly inserted (Section V-B1), the reported differences may be within run-to-run variability. Additionally, the closest adaptive baselines discussed in the Introduction—Adaptive KalmanNet [28], EM-KalmanNet [29], and MAML-KalmanNet [30]—are never benchmarked. The abstract's claim of 'consistently outperforming existing learning-based filters' is therefore not yet fully supported. Please add repeated-seed statistics and at least one of these adaptive baselines.
  4. [Section V-A2/Table I and Section II-B] CPDNet is trained and tested on the same type of parameter variations (disjoint trajectories but shared change types), so Table I only demonstrates in-distribution detection. Section III-B1 further acknowledges that slow drifts much slower than the window δ may be missed. The paper should test CPDNet on held-out change types (e.g., a change in Q when trained on R changes) and on slower drifts to substantiate the 'change-aware' claim beyond the trained regimes.
  5. [Section IV-C and Section V-C (last paragraph)] The paper explicitly limits its scope to single-change, finite-horizon scenarios and states that significant structural changes would require system identification. These are honest limitations, but they directly qualify the abstract's broad claims of handling 'temporal variations' and 'model mismatches' in general. Please either temper the claims to the tested regime or add experiments with repeated/multiple changes to support the stated scope.
minor comments (6)
  1. [Section V-B2] The sentence 'retaining over 55% of the performance gap between hybrid-driven and optimal methods' is ambiguous; please clarify whether 55% of the gap is retained or closed, and give the computation.
  2. [Section V-A1] The paragraph describing the constant-acceleration model is duplicated; remove the repetition.
  3. [Table I] The heading 'MSE (dB)' for a comparison of values in [0,1] is nonstandard; consider reporting mean absolute error or similar, with dB used only if clearly defined.
  4. [Section V-B1] The notation '1/r^2 [dB]' is confusing: state explicitly that R = r^2 I and that dB refers to 10 log10(1/r^2).
  5. [Figures 3 and 4] The bottom panels plot both state and observation values with labels 'Estimation y'/'True y'; clarify the units and what is being compared (state vs output).
  6. [Algorithm 2, line 8] After the update step on line 7, the Kalman gain should be recomputed with the updated θ; the pseudocode is ambiguous as written.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: CPDNet is supervised by external state-error labels and the adaptation loss is an independent innovation-based proxy.

full rationale

The paper's derivation chain is self-contained rather than circular. CPDNet is trained as a supervised detector: its input features are a nonlinear function of the innovation norm (Eq. 10), while its training target is a nonlinear function of the true state-estimation error (Eq. 13). These are different quantities, and the target is anchored to ground-truth state trajectories in the offline dataset D_v, not constructed from the same innovation signal in a way that would make the prediction tautological. The online adaptation loss (Eq. 16) minimizes the innovation norm plus regularization; this is a heuristic proxy taken from prior work [50], not the evaluation metric (Eq. 23, state MSE). The paper explicitly calls the resulting update a 'CPDNet-gated heuristic adaptation magnitude' and does not claim the loss is theoretically equivalent to state error. The self-citations to KalmanNet [19] and to the unsupervised innovation-based loss [50] are normal references to published, externally developed results; they do not smuggle in the paper's conclusion. The CPDNet evaluation in Table I compares its output to the separately computed KalmanNet error on held-out change scenarios, which is a legitimate generalization test, not a fitted-input prediction. The manuscript also includes honest limitations (e.g., the detection window may miss slow drift, and repeated changes are left for future work), which further indicates the authors are not disguising assumptions as derivations. No step in the paper reduces, by construction or by self-citation, to its own inputs.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

The framework rests on several unproved modeling assumptions: innovation magnitude proxies filter error, minimizing it helps state estimation, f and h remain reliable, and test changes resemble training changes. The method also has at least six hyperparameters not reported in the paper. These are load-bearing for the empirical claims.

free parameters (7)
  • gamma (feature/label scaling) = not reported
    Appears in Eqs. (10) and (13); scales innovation norm and error into [0,1]. No value or sensitivity analysis in the text.
  • thresh (CPDNet trigger threshold) = not reported
    In Eq. (15); determines when adaptation is zero vs active. No value reported; performance likely sensitive to it.
  • epsilon (max online learning rate) = not reported
    In Eq. (15); upper bound on eta_t. No value reported.
  • rho (regularization coefficient) = not reported
    In Eq. (16); balances innovation loss vs parameter drift. No value reported.
  • delta (CPDNet window length) = not reported
    In Eq. (11); controls temporal receptive field and detection delay. No value reported.
  • tau_max (truncated BPTT horizon) = not reported
    Algorithm 2 / Section IV-A2; controls update locality and compute. No value reported.
  • CPDNet architecture hyperparameters (RNN hidden size, layers) = not reported
    Neural module structure in Section III-B2; sizes not given.
assumptions (4)
  • domain assumption The l2 norm of the innovation process is a reliable indicator of KalmanNet misalignment.
    Motivates CPDNet features (Eq. 10) and the unsupervised loss (Eq. 16); no proof or isolated validation. Location: Section III-A and IV-A2.
  • domain assumption Minimizing the squared innovation norm with one gradient step improves state estimation after a distribution shift.
    Adaptation rule Eq. (17) is built on this; the paper explicitly calls the result a heuristic rather than a guaranteed optimal update. Location: Section IV-A1/2.
  • domain assumption The state-transition f(.) and observation h(.) functions are known well enough to compute meaningful innovations.
    Algorithm 2 requires f,h at every step; the paper admits significant structural changes would need system identification. Location: Section IV-C.
  • domain assumption The change types encountered at test time belong to the same family as those in the nonstationary training set D_v.
    CPDNet is trained on D_v with specific parameter variations and tested on the same types of variations. Location: Section II-B.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Change-Aware Self-Adaptive AI-Aided Kalman Filters With Neural Change Point Detection." pith.science (2026). https://pith.science/paper/4Y7J2532

@misc{pith2026260713387,
  author       = {Pith},
  title        = {Pith review of: Change-Aware Self-Adaptive AI-Aided Kalman Filters With Neural Change Point Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4Y7J2532}},
  note         = {Machine review of arXiv:2607.13387}
}
read the original abstract

Reliable state estimation in dynamical systems is often challenged by model mismatches, unknown noise statistics, and temporal variations. While AI-aided Kalman filters such as KalmanNet leverage deep learning to enhance classical estimation, they remain vulnerable to distribution shifts and lack mechanisms for autonomous adaptation. This work introduces Change-Aware Self-Adaptive KalmanNet (CASA-KalmanNet), an online adaptation framework that integrates a dedicated neural module, termed CPDNet, to monitor the interpretable internal features of KalmanNet and provide soft indicators of reliability degradation. These indicators dynamically regulate an online learning process, enabling data-efficient and timely adaptation to both abrupt and gradual changes in the system without requiring additional state labels from the changed regime. Numerical experiments on linear and nonlinear state-space models show that CASA-KalmanNet consistently outperforms existing learning-based filters under model mismatch, while approaching the accuracy of optimal classical methods with full domain knowledge.

Figures

Figures reproduced from arXiv: 2607.13387 by the authors.

Figure 1
Figure 1. Overall architecture of CPDNet framework. KalmanNet’s [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. CASA-KalmanNet illustration.The framework combines KalmanNet with CPDNet-based online adaptation, where CPDNet extracts statistical features from the innovation to adjust the learning rate for updating KalmanNet parameters. Z −1 denotes the unit delay operator. γupd > 0, where len(·) applied to a vector returns its length. Using this introduced notation, the excessive complexity becomes ∆C = C CPD inf + Pr(ηt > 0) ·… view at source ↗
Figure 3
Figure 3. Performance of CPDNet under gradual changes in [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Performance of CPDNet under λ = 1.01 change in state transition matrix F, with the same plotting format as in [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Linear model result on H rotate after change −10 −5 0 5 −30 −20 −10 0 10 20 30 MSE(dB) (dB) KF KalmanNet CASA-KalmanNet Unsupervised-KalmanNet EM-KF [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Linear model result on Q change gradually after change 3) Partial information results: We proceed to evaluate CASA-KalmanNet and the baseline methods under partial information settings, where the generative parameters are only partially known to the filtering algorithm…
Figure 7
Figure 7. Figure 7: Position MSE performance under partial information [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 9
Figure 9. Figure 9: Performance comparison on the Lorenz attractor under [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 11
Figure 11. Figure 11: Performance comparison on the Lorenz attractor under [PITH_FULL_IMAGE:figures/full_fig_p012_11.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. ARC: Augmented-Rank Conformalization for Changepoint Localization --- Finite-Sample Validity and Distribution-Robust Efficiency

    stat.ML 2026-08 conditional novelty 6.0 of 10

    Rank-based conformal scores make changepoint localization sets finite-sample valid for any frozen weights and exactly invariant to monotone data transforms, transferring certified set lengths across the entire monotone orbit.

Reference graph

Works this paper leans on

53 extracted references · 4 linked inside Pith · cited by 1 Pith paper

  1. [28]

    Adaptive KalmanNet: Data-driven Kalman filter with fast adaptation,

    X. Ni, G. Revach, and N. Shlezinger, “Adaptive KalmanNet: Data-driven Kalman filter with fast adaptation,” inProc. IEEE Int. Conf. Acoust. Speech Signal Process., 2024, pp. 5970–5974

    2024

  2. [29]

    EM-Kalmannet: AI-aided kalman tracking in partially known time-varying state-space models,

    O. Cohen, X. Ni, N. Shlezinger, and T. Routtenberg, “EM-Kalmannet: AI-aided kalman tracking in partially known time-varying state-space models,” inIEEE Sensor Array and Multichannel Signal Processing Workshop (SAM), 2026

    2026

  3. [30]

    MAML- KalmanNet: A neural network-assisted Kalman filter based on model-agnostic meta-learning,

    S. Chen, Y . Zheng, D. Lin, P. Cai, Y . Xiao, and S. Wang, “MAML- KalmanNet: A neural network-assisted Kalman filter based on model-agnostic meta-learning,”IEEE Trans. Signal Process., vol. 73, pp. 988–1003, 2025

    2025

  4. [1]

    Bar-Shalom, X

    Y . Bar-Shalom, X. R. Li, and T. Kirubarajan,Estimation With Applications to Tracking and Navigation: Theory Algorithms and Software. John Wiley & Sons, 2004

    2004

  5. [2]

    Tracking targets using adaptive kalman filter- ing,

    P.-O. Gutman and M. Velger, “Tracking targets using adaptive kalman filter- ing,”IEEE Trans. Aerosp. Electron. Syst., vol. 26, no. 5, pp. 691–699, 1990

    1990

  6. [3]

    Durbin and S

    J. Durbin and S. J. Koopman,Time Series Analysis by State Space Methods. Oxford University Press, 2012

    2012

  7. [4]

    The Kalman filter,

    S. Gannot and A. Yeredor, “The Kalman filter,”Springer Handbook of Speech Processing, pp. 135–160, 2008

    2008

  8. [5]

    A new approach to linear filtering and prediction problems,

    R. E. Kalman, “A new approach to linear filtering and prediction problems,” J. Basic Eng., vol. 82, no. 1, pp. 35–45, 1960

    1960

Show all 53 references
  1. [6]

    The Kalman filter-its recognition and development for aerospace applications,

    S. F. Schmidt, “The Kalman filter-its recognition and development for aerospace applications,”J. Guid. Control Dyn., vol. 4, no. 1, pp. 4–7, 1981

    1981

  2. [7]

    Unscented filtering and nonlinear estimation,

    S. J. Julier and J. K. Uhlmann, “Unscented filtering and nonlinear estimation,”Proc. IEEE, vol. 92, no. 3, pp. 401–422, 2004

    2004

  3. [8]

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

    N. Shlezinger, G. Revach, A. Ghosh, S. Chatterjee, S. Tang, T. Imbiriba, J. Dunik, O. Straka, P. Closas, and Y . C. Eldar, “Artificial intelligence-aided Kalman filters: AI-augmented designs for Kalman-type algorithms,”IEEE Signal Process. Mag., 2025, early access

    2025

  4. [9]

    Model-based deep learning,

    N. Shlezinger, J. Whang, Y . C. Eldar, and A. G. Dimakis, “Model-based deep learning,”Proc. IEEE, vol. 111, no. 5, pp. 465–499, 2023

    2023

  5. [10]

    Augmented physics-based machine learning for navigation and tracking,

    T. Imbiriba, O. Straka, J. Duník, and P. Closas, “Augmented physics-based machine learning for navigation and tracking,”IEEE Trans. Aerosp. Electron. Syst., vol. 60, no. 3, pp. 2692–2704, 2024

    2024

  6. [11]

    Learning nonlinear state–space models using autoencoders,

    D. Masti and A. Bemporad, “Learning nonlinear state–space models using autoencoders,”Automatica, vol. 129, p. 5–1098, 2021

    2021

  7. [12]

    NNAKF: A neural network adapted Kalman filter for target tracking,

    S. Jouaber, S. Bonnabel, S. Velasco-Forero, and M. Pilte, “NNAKF: A neural network adapted Kalman filter for target tracking,” inProc. IEEE Int. Conf. Acoust. Speech Signal Process., 2021, pp. 4075–4079

    2021

  8. [13]

    EKFNet: Learning system noise covariance parameters for nonlinear tracking,

    L. Xu and R. Niu, “EKFNet: Learning system noise covariance parameters for nonlinear tracking,”IEEE Trans. Signal Process., vol. 72, pp. 3139–3152, 2024

    2024

  9. [14]

    Adaptive Kalman-informed transformer,

    N. Cohen and I. Klein, “Adaptive Kalman-informed transformer,”Eng. Appl. Artif. Intell., vol. 146, p. 110221, 2025

    2025

  10. [15]

    Combining generative and discriminative models for hybrid inference,

    V . G. Satorras, Z. Akata, and M. Welling, “Combining generative and discriminative models for hybrid inference,” inProc. Adv. Neural Inf. Process., 2019, pp. 13 802–13 812

    2019

  11. [16]

    Neural augmentation of Kalman filter with hypernetwork for channel tracking,

    K. Pratik, R. A. Amjad, A. Behboodi, J. B. Soriaga, and M. Welling, “Neural augmentation of Kalman filter with hypernetwork for channel tracking,” inProc. IEEE Glob. Commun. Conf., 2021

    2021

  12. [17]

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

    A. Ghosh, A. Honoré, and S. Chatterjee, “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

  13. [18]

    DeepBayes—an estimator for parameter estimation in stochastic nonlinear dynamical models,

    A. Ghosh, M. Abdalmoaty, S. Chatterjee, and H. Hjalmarsson, “DeepBayes—an estimator for parameter estimation in stochastic nonlinear dynamical models,”Automatica, vol. 159, p. 111327, 2024

    2024

  14. [19]

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

    G. Revach, N. Shlezinger, X. Ni, A. L. Escoriza, R. J. Van Sloun, and Y . C. Eldar, “KalmanNet: Neural network aided Kalman filtering for partially known dynamics,”IEEE Trans. Signal Process., vol. 70, pp. 1532–1547, 2022

    2022

  15. [20]

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

    G. Revach, X. Ni, N. Shlezinger, R. J. van Sloun, and Y . C. Eldar, “RTSNet: Learning to smooth in partially known state-space models,” IEEE Trans. Signal Process., vol. 71, pp. 4441–4456, 2023

    2023

  16. [21]

    Split-KalmanNet: A robust model-based deep learning approach for state estimation,

    G. Choi, J. Park, N. Shlezinger, Y . C. Eldar, and N. Lee, “Split-KalmanNet: A robust model-based deep learning approach for state estimation,”IEEE Trans. Veh. Technol., vol. 72, no. 9, pp. 12 326–12 331, 2023

    2023

  17. [22]

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

    I. Buchnik, G. Revach, D. Steger, R. J. Van Sloun, T. Routtenberg, and N. Shlezinger, “Latent-KalmanNet: Learned Kalman filtering for tracking from high-dimensional signals,”IEEE Trans. Signal Process., vol. 72, pp. 352–367, 2023

    2023

  18. [23]

    Nonlinear Kalman filtering based on self-attention mechanism and lattice trajectory piecewise linear approximation,

    J. Wang, X. Geng, and J. Xu, “Nonlinear Kalman filtering based on self-attention mechanism and lattice trajectory piecewise linear approximation,”arXiv preprint arXiv:2404.03915, 2024

    2024 arXiv

  19. [24]

    GSP-KalmanNet: Tracking graph signals via neural-aided Kalman filtering,

    I. Buchnik, G. Sagi, N. Leinwand, Y . Loya, N. Shlezinger, and T. Routten- berg, “GSP-KalmanNet: Tracking graph signals via neural-aided Kalman filtering,”IEEE Trans. Signal Process., vol. 72, pp. 3700–3716, 2024

    2024

  20. [25]

    State estimation with 1-bit observations and imperfect models: Bussgang meets Kalman in neural networks,

    C. Jung, T. Ha, H. Kim, and J. Park, “State estimation with 1-bit observations and imperfect models: Bussgang meets Kalman in neural networks,”arXiv preprint arXiv:2507.17284, 2025

    2025 arXiv

  21. [26]

    Discriminative and generative learning for linear estimation of random signals [lecture notes],

    N. Shlezinger and T. Routtenberg, “Discriminative and generative learning for linear estimation of random signals [lecture notes],”IEEE Signal Process. Mag., vol. 40, no. 6, pp. 75–82, 2023

    2023

  22. [27]

    Towards out- of-distribution generalization: A survey,

    J. Liu, Z. Shen, Y . He, X. Zhang, R. Xu, H. Yu, and P. Cui, “Towards out- of-distribution generalization: A survey,”arXiv preprint arXiv:2108.13624, 2021

    2021 arXiv

  23. [31]

    Approaches to adaptive filtering,

    R. Mehra, “Approaches to adaptive filtering,”IEEE Trans. Autom. Control, vol. 17, no. 5, pp. 693–698, 1972

    1972

  24. [32]

    An approach to time series smoothing and forecasting using the EM algorithm,

    R. H. Shumway and D. S. Stoffer, “An approach to time series smoothing and forecasting using the EM algorithm,”J. Time Ser. Anal., vol. 3, no. 4, pp. 253–264, 1982

    1982

  25. [33]

    Variational adaptive Kalman filter with Gaussian-Inverse-Wishart mixture distribution,

    Y . Huang, Y . Zhang, P. Shi, and J. Chambers, “Variational adaptive Kalman filter with Gaussian-Inverse-Wishart mixture distribution,”IEEE Trans. Automat. Contr., vol. 66, no. 4, pp. 1786–1793, 2021

    2021

  26. [34]

    A novel robust Kalman filtering framework based on normal-skew mixture distribution,

    M. Bai, Y . Huang, B. Chen, and Y . Zhang, “A novel robust Kalman filtering framework based on normal-skew mixture distribution,”IEEE Trans. Syst. Man Cybern. Syst., vol. 52, no. 11, pp. 6789–6805, 2022

    2022

  27. [35]

    Distributed generalized minimum error entropy unscented Kalman filter under hybrid attacks without prior knowledge,

    J. He, G. Wang, Z. Feng, S. Zhong, P. Zhang, and B. Peng, “Distributed generalized minimum error entropy unscented Kalman filter under hybrid attacks without prior knowledge,”IEEE Trans. Instrum. Meas., vol. 74, pp. 1–12, 2025

    2025

  28. [36]

    Variational nonlinear Kalman filtering with unknown process noise covariance,

    H. Lan, J. Hu, Z. Wang, and Q. Cheng, “Variational nonlinear Kalman filtering with unknown process noise covariance,”IEEE Trans. Aerosp. Electron. Syst., vol. 59, no. 6, pp. 9177–9190, 2023

    2023

  29. [37]

    Asynchronous online adaptation via modular drift detection for deep receivers,

    N. Uzlaner, T. Raviv, N. Shlezinger, and K. Todros, “Asynchronous online adaptation via modular drift detection for deep receivers,”IEEE Trans. Wireless Commun., vol. 24, no. 5, pp. 4454–4468, 2025

    2025

  30. [38]

    University of Michigan north campus long-term vision and lidar dataset,

    N. Carlevaris-Bianco, A. K. Ushani, and R. M. Eustice, “University of Michigan north campus long-term vision and lidar dataset,”Int. J. Robot. Res., vol. 35, no. 9, pp. 1023–1035, 2015

    2015

  31. [39]

    Cubature Kalman filters,

    I. Arasaratnam and S. Haykin, “Cubature Kalman filters,”IEEE Trans. Autom. Control, vol. 54, no. 6, pp. 1254–1269, 2009

    2009

  32. [40]

    Model-based deep learning,

    N. Shlezinger and Y . C. Eldar, “Model-based deep learning,”Found. Trends Signal Process., vol. 17, no. 4, pp. 291–416, 2023

    2023

  33. [41]

    A survey of methods for time series change point detection,

    S. Aminikhanghahi and D. J. Cook, “A survey of methods for time series change point detection,”Knowl. Inf. Syst., vol. 51, no. 2, pp. 339–367, 2017

    2017

  34. [42]

    Bayesian online change point detection,

    R. P. Adams and D. J. MacKay, “Bayesian online change point detection,” arXiv preprint arXiv:0710.3742, 2007

    2007 arXiv

  35. [43]

    Restarted Bayesian online change-point detector achieves optimal detection delay,

    R. Alami, O. Maillard, and R. Féraud, “Restarted Bayesian online change-point detector achieves optimal detection delay,” inProc. Int. Conf. Mach. Learn.PMLR, 2020, pp. 211–221

    2020

  36. [44]

    Kernel change-point analysis,

    Z. Harchaoui, E. Moulines, and F. Bach, “Kernel change-point analysis,” Proc. Adv. Neural Inf. Process., vol. 21, 2008

    2008

  37. [45]

    Practical and powerful kernel-based change-point detection,

    H. Song and H. Chen, “Practical and powerful kernel-based change-point detection,”IEEE Trans. Signal Process., vol. 72, pp. 5174–5186, 2024

    2024

  38. [46]

    Learning under concept drift: A review,

    J. Lu, A. Liu, F. Dong, F. Gu, J. Gama, and G. Zhang, “Learning under concept drift: A review,”IEEE Trans. Knowl. Data Eng., vol. 31, no. 12, pp. 2346–2363, 2018

    2018

  39. [47]

    Learning with drift detection,

    J. Gama, P. Medas, G. Castillo, and P. Rodrigues, “Learning with drift detection,” inBrazilian Symposium on Artificial Intelligence. Springer, 2004, pp. 286–295

    2004

  40. [48]

    Concept drift detection for deep learning aided receivers in dynamic channels,

    N. Uzlaner, T. Raviv, N. Shlezinger, and K. Todros, “Concept drift detection for deep learning aided receivers in dynamic channels,” inProc. IEEE Workshop Signal Process. Adv. Wirel., 2024, pp. 371–375

    2024

  41. [49]

    Adversarial learning targeting deep neural network classification: A comprehensive review of defenses against attacks,

    D. J. Miller, Z. Xiang, and G. Kesidis, “Adversarial learning targeting deep neural network classification: A comprehensive review of defenses against attacks,”Proc. IEEE, vol. 108, no. 3, pp. 402–433, 2020

    2020

  42. [50]

    Unsupervised learned Kalman filtering,

    G. Revach, N. Shlezinger, T. Locher, X. Ni, R. J. van Sloun, and Y . C. Eldar, “Unsupervised learned Kalman filtering,” inProc. Eur. Signal Process. Conf.IEEE, 2022, pp. 1571–1575

    2022

  43. [51]

    Training recurrent neural networks,

    I. Sutskever, “Training recurrent neural networks,” Ph.D. dissertation, 2013

    2013

  44. [52]

    Memory-efficient backpropagation through time,

    A. Gruslys, R. Munos, I. Danihelka, M. Lanctot, and A. Graves, “Memory-efficient backpropagation through time,”Proc. Adv. Neural Inf. Process., vol. 29, 2016. 15

    2016

  45. [53]

    Uncertainty quantification in deep learning based Kalman filters,

    Y . Dahan, G. Revach, J. Dunik, and N. Shlezinger, “Uncertainty quantification in deep learning based Kalman filters,” inProc. IEEE Int. Conf. Acoust. Speech Signal Process., 2024, pp. 13 121–13 125

    2024

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.