pith. machine review for the scientific record. sign in

arxiv: 2605.05520 · v1 · submitted 2026-05-06 · 💻 cs.LG · stat.AP· stat.ML

Recognition: unknown

Bayesian Rain Field Reconstruction using Commercial Microwave Links and Diffusion Model Priors

Albina Ilina, Badr Moufad, Eric Moulines, Hagit Messer, Hai Victor Habi, Salem Lahlou, Yazid Janati

Pith reviewed 2026-05-08 16:20 UTC · model grok-4.3

classification 💻 cs.LG stat.APstat.ML
keywords rainfall reconstructioncommercial microwave linksdiffusion modelsBayesian inverse problemsspatial priorsprecipitation estimationinverse problems
0
0 comments X

The pith

Diffusion models as priors in a Bayesian inverse problem improve rainfall field reconstruction from commercial microwave link measurements.

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

The paper reframes reconstructing rainfall fields from commercial microwave link data as a Bayesian inverse problem. It uses pre-trained diffusion models to serve as high-fidelity priors for the spatial distribution of rain. This approach allows for training-free sampling of the posterior distribution using methods like plug-and-play and sequential Monte Carlo. The result is better preservation of rainfall statistics and improved accuracy over traditional methods, especially in varied precipitation conditions. A sympathetic reader would care because accurate rain mapping supports better weather prediction, flood management, and agricultural planning.

Core claim

Viewing rain field reconstruction as a Bayesian inverse problem with diffusion models as spatial priors enables training-free posterior sampling. Diffusion models better preserve key rainfall statistics than censored Gaussian processes. This framework leads to consistent improvements in reconstruction accuracy on both synthetic and real-world datasets compared to established CML-based baselines.

What carries the argument

Bayesian inverse problem formulation with diffusion model priors for path-integrated attenuation measurements from commercial microwave links.

If this is right

  • Better reconstruction under heterogeneous precipitation conditions.
  • Training-free application of advanced priors without domain-specific fine-tuning.
  • Potential for integration with various sampling methods including Plug-and-Play, Sequential Monte Carlo, and Replica Exchange.
  • Improved ground-level rainfall estimates from dense but line-integrated sensor networks.

Where Pith is reading between the lines

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

  • This could extend to real-time nowcasting systems if sampling is efficient enough.
  • Connecting to other inverse problems in environmental sensing where diffusion priors might apply.
  • If validated more broadly, it suggests pre-trained generative models can substitute for hand-crafted priors in geophysical inverse problems.

Load-bearing premise

Pre-trained diffusion models without adaptation accurately capture the spatial statistics of rainfall, and the forward model of line-integrated attenuation holds for heterogeneous precipitation.

What would settle it

A direct comparison on real-world data where ground truth from rain gauges shows no improvement in metrics like RMSE or correlation over baselines would falsify the claim of consistent improvements.

Figures

Figures reproduced from arXiv: 2605.05520 by Albina Ilina, Badr Moufad, Eric Moulines, Hagit Messer, Hai Victor Habi, Salem Lahlou, Yazid Janati.

Figure 1
Figure 1. Figure 1: Comparison of rain field statistics between the refer￾ence samples and generated samples using the DM prior. The histograms show densities and are built using 5000 samples. exists a rich literature on stochastic precipitation model￾ing and geostatistical reconstruction (Wilks & Wilby, 1999; Wheater et al., 2000; Yang et al., 2005). A classical and widely used statistical abstraction models this mixed behav… view at source ↗
Figure 2
Figure 2. Figure 2: Examples of generated rain fields using the prior DM compared with reference rain fields. A total of 5,000 samples were generated. In each subfigure, on the left, samples are randomly drawn from the datasets; on the right, they are drawn from the top 90% wet samples by cumulative rain. clean sample x0 in the above transition with the output of a neural network (t, xt) 7→ Dθ t (xt) parameterized by θ. This … view at source ↗
Figure 3
Figure 3. Figure 3: Comparison between the reconstructions of the baselines on the inverse problem with diffusion prior on the setting of GP. The y-axis shows intensities while the x-axis represents a one-dimensional grid of the [−5, 5] interval. The dashed horizontal lines depict values of the integral over the observed intervals. likelihood of the inverse problem is p(y | x0) ∝ exp{−1 2 ∥y − M(x0)∥ 2 Σ−2 }, (5) with ∥v∥ 2 Σ… view at source ↗
Figure 4
Figure 4. Figure 4: Comparisons of rain field reconstructions on real CMLs links from OpenMRG. We depict the network of CMLs in red view at source ↗
Figure 5
Figure 5. Figure 5: Power-law parameters of the CML as a function of link’s length and the frequency on the OpenMRG dataset. C.1. Meteorological baselines IDW. We implement inverse-distance weighting (IDW) following Eshel et al. (2021); Shepard (1968). Each CML link observation is represented as a point measurement located at the link midpoint. For every grid location s, we compute a normalized weighted average using p = 2 an… view at source ↗
Figure 6
Figure 6. Figure 6: Illustration of a line sensor on a 6 × 4 grid. The blue dots define the line sensor whereas the red dots mark its intersections with the two-dimensional grid. The color intensity of each cell is proportional to the intersection length. Considered convention. In our setting, the radar maps use the origin (− 1 2 , − 1 2 ) and unit spacing in both directions, i.e., ∆x = ∆y = 1. Indexing cells by the integer c… view at source ↗
Figure 7
Figure 7. Figure 7: Extended results of view at source ↗
Figure 8
Figure 8. Figure 8: Qualitative comparisons of rain field reconstructions on real CMLs links from OpenMRG dataset on three reconstruction tasks. The network of CMLs is depicted in red. 21 view at source ↗
Figure 9
Figure 9. Figure 9: Another set of qualitative comparisons of rain field reconstructions on real CMLs links from OpenMRG dataset on three reconstruction tasks. The network of CMLs is depicted in red. 22 view at source ↗
read the original abstract

Commercial Microwave Links (CMLs) offer dense spatial coverage for rainfall sensing but produce path-integrated measurements that make accurate ground-level reconstruction challenging. Existing methods typically oversimplify CMLs as point sensors and neglect line integration relating rainfall to signal attenuation, resulting in degraded performance under heterogeneous precipitation. In this work, we view rain field reconstruction as a Bayesian inverse problem with Diffusion Models (DMs) as high-fidelity spatial priors. We show that diffusion models better preserve key rainfall statistics compared to censored Gaussian processes. Framing rainfall estimation as a Bayesian inverse problem with a DM prior enables training-free posterior sampling using a broad family of methods, including Plug-and-Play, Sequential Monte Carlo, and Replica Exchange methods. Experiments on synthetic and real-world datasets demonstrate consistent improvements over established CML-based reconstruction baselines.

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

2 major / 2 minor

Summary. The manuscript frames rainfall field reconstruction from commercial microwave link (CML) path-integrated attenuation measurements as a Bayesian inverse problem. It uses pre-trained diffusion models as spatial priors and applies training-free posterior sampling via Plug-and-Play, Sequential Monte Carlo, and Replica Exchange methods. The central claims are that diffusion model priors better preserve key rainfall statistics (e.g., intermittency and heavy tails) than censored Gaussian processes and that the approach yields consistent improvements over established CML-based baselines on both synthetic and real-world datasets.

Significance. If the claims hold, the work would be significant for demonstrating the utility of off-the-shelf diffusion model priors in Bayesian inverse problems involving non-Gaussian spatial fields, particularly for environmental sensing applications. A notable strength is the reliance on standard sampling techniques applied to a new modality without introducing free parameters or circular derivations. The emphasis on statistic preservation addresses a known shortcoming of Gaussian process methods in precipitation modeling.

major comments (2)
  1. [§4 (Experiments)] §4 (Experiments): The abstract and results section claim consistent improvements over baselines and better preservation of rainfall statistics than censored GPs, but provide no quantitative details on exact error metrics (e.g., RMSE, MAE), specific baseline implementations, or ablation studies on sampling methods. This leaves the magnitude and robustness of the reported gains difficult to assess and undermines evaluation of the central claim.
  2. [§3 (Method)] §3 (Method) and real-world dataset results: The load-bearing assumption that pre-trained (training-free) diffusion models faithfully capture rainfall intermittency, heavy-tailed statistics, and spatial structure is not supported by direct quantitative checks. No comparisons of DM prior samples against observed variograms, exceedance probabilities, or wet/dry area fractions from the real dataset are reported, which directly risks the validity of the superiority claim over censored GPs.
minor comments (2)
  1. [§2] The forward model of line-integrated attenuation is described at a high level; adding an explicit equation or pseudocode in §2 would improve clarity for readers unfamiliar with CML physics.
  2. [Figures] Figure captions and axis labels in the results section could more explicitly indicate which sampling method (PnP, SMC, or Replica Exchange) is shown in each panel to aid reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment point by point below and have revised the manuscript to incorporate additional quantitative details and validations where the original presentation was insufficient.

read point-by-point responses

  1. Referee: [§4 (Experiments)] The abstract and results section claim consistent improvements over baselines and better preservation of rainfall statistics than censored GPs, but provide no quantitative details on exact error metrics (e.g., RMSE, MAE), specific baseline implementations, or ablation studies on sampling methods. This leaves the magnitude and robustness of the reported gains difficult to assess and undermines evaluation of the central claim.

    Authors: We agree that the original manuscript would benefit from more explicit numerical reporting. While Section 4 includes visual comparisons and aggregate performance indicators, tabulated error metrics, precise baseline specifications, and sampling ablations were not provided. In the revised version we have added a summary table reporting RMSE, MAE, and Pearson correlation for all methods on both synthetic and real datasets, along with a detailed description of baseline implementations (including the censored GP setup) and an ablation comparing Plug-and-Play, SMC, and Replica Exchange sampling. revision: yes

  2. Referee: [§3 (Method)] and real-world dataset results: The load-bearing assumption that pre-trained (training-free) diffusion models faithfully capture rainfall intermittency, heavy-tailed statistics, and spatial structure is not supported by direct quantitative checks. No comparisons of DM prior samples against observed variograms, exceedance probabilities, or wet/dry area fractions from the real dataset are reported, which directly risks the validity of the superiority claim over censored GPs.

    Authors: We acknowledge that direct quantitative validation of the diffusion prior against real-data statistics is important for supporting the central claims. The manuscript demonstrates advantages through improved posterior reconstructions, but does not include standalone prior-sample diagnostics. In the revised manuscript we have added a new subsection with explicit comparisons of diffusion-model samples to the real dataset (and to the censored GP) using variograms, exceedance probabilities, and wet/dry area fractions; these confirm that the DM prior better reproduces the observed non-Gaussian spatial structure. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation applies standard Bayesian methods to new modality

full rationale

The paper frames rain field reconstruction as a Bayesian inverse problem, adopts pre-trained diffusion models as spatial priors (without domain-specific fine-tuning or fitting in this work), and invokes established training-free posterior sampling algorithms (Plug-and-Play, SMC, Replica Exchange). All central claims rest on experimental validation against baselines on synthetic and real CML datasets rather than any self-referential definitions, fitted parameters renamed as predictions, or load-bearing self-citations. No equation or step reduces a claimed result to its own inputs by construction; the approach is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The abstract provides no explicit free parameters, axioms, or invented entities; the method assumes standard diffusion model training on rainfall-like fields and standard Bayesian inverse problem formulations.

axioms (2)
  • domain assumption Diffusion models trained on general spatial data serve as suitable high-fidelity priors for rainfall fields
    Invoked when stating that DMs better preserve key rainfall statistics than censored Gaussian processes.
  • domain assumption The line integration relating rainfall to signal attenuation can be accurately modeled in the forward operator
    Central to framing the problem as a Bayesian inverse problem.

pith-pipeline@v0.9.0 · 5454 in / 1296 out tokens · 36013 ms · 2026-05-08T16:20:36.312900+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

183 extracted references · 28 canonical work pages · 8 internal anchors

  1. [1]

    AMC , volume=

    Bayesian bounds for parameter estimation and nonlinear filtering/tracking , author=. AMC , volume=

  2. [2]

    2014 , publisher=

    Understanding machine learning: From theory to algorithms , author=. 2014 , publisher=

    2014

  3. [3]

    IEEE Transactions on Acoustics, Speech, and Signal Processing , volume=

    Array shape calibration using sources in unknown locations--Part I: Far-field sources , author=. IEEE Transactions on Acoustics, Speech, and Signal Processing , volume=. 1987 , publisher=

    1987

  4. [4]

    1993 , publisher=

    Fundamentals of statistical signal processing , author=. 1993 , publisher=

    1993

  5. [5]

    Advances in Neural Information Processing Systems , pages=

    Masked autoregressive flow for density estimation , author=. Advances in Neural Information Processing Systems , pages=

  6. [6]

    Applications of the van Trees inequality: a Bayesian Cram

    Gill, Richard D and Levit, Boris Y and others , journal=. Applications of the van Trees inequality: a Bayesian Cram. 1995 , publisher=

    1995

  7. [7]

    Advances in neural information processing systems , pages=

    Generative adversarial nets , author=. Advances in neural information processing systems , pages=

  8. [8]

    The Gaussian data assumption leads to the largest Cram

    Stoica, Petre and Babu, Prabhu , journal=. The Gaussian data assumption leads to the largest Cram. 2011 , publisher=

    2011

  9. [9]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Noise flow: Noise modeling with conditional normalizing flows , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  10. [10]

    Proceedings of the International Conference on Learning Representations (ICLR) , volume=

    Auto-encoding variational bayes , author=. Proceedings of the International Conference on Learning Representations (ICLR) , volume=

  11. [11]

    Conditional Generative Adversarial Nets

    Conditional generative adversarial nets , author=. arXiv preprint arXiv:1411.1784 , year=

    work page internal anchor Pith review arXiv

  12. [12]

    Proceedings of the 31st International Conference on Machine Learning , pages =

    Stochastic Backpropagation and Approximate Inference in Deep Generative Models , author =. Proceedings of the 31st International Conference on Machine Learning , pages =. 2014 , editor =

    2014

  13. [13]

    Kingma and Jimmy Ba , editor =

    Diederik P. Kingma and Jimmy Ba , editor =. Adam:. 3rd International Conference on Learning Representations,. 2015 , url =

    2015

  14. [14]

    2017 IEEE International Conference on Computer Vision (ICCV) , title=

    X. 2017 IEEE International Conference on Computer Vision (ICCV) , title=. 2017 , volume=

    2017

  15. [15]

    Rectifier nonlinearities improve neural network acoustic models , author=

  16. [16]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Deep residual learning for image recognition , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  17. [17]

    2017 , publisher=

    Deep learning , author=. 2017 , publisher=

    2017

  18. [18]

    Proceedings of the European conference on computer vision (ECCV) , pages=

    Cbam: Convolutional block attention module , author=. Proceedings of the European conference on computer vision (ECCV) , pages=

  19. [19]

    International conference on machine learning , pages=

    Autoencoding beyond pixels using a learned similarity metric , author=. International conference on machine learning , pages=. 2016 , organization=

    2016

  20. [20]

    arXiv preprint arXiv:2011.08848 , year=

    Deep Networks for Direction-of-Arrival Estimation in Low SNR , author=. arXiv preprint arXiv:2011.08848 , year=

    work page arXiv 2011

  21. [21]

    1992 , publisher=

    Array signal processing: concepts and techniques , author=. 1992 , publisher=

    1992

  22. [22]

    2004 , publisher=

    Optimum array processing: Part IV of detection, estimation, and modulation theory , author=. 2004 , publisher=

    2004

  23. [23]

    , journal=

    Liu, Zhang-Meng and Zhang, Chenwei and Yu, Philip S. , journal=. Direction-of-Arrival Estimation Based on Deep Neural Networks With Robustness to Array Imperfections , year=

  24. [24]

    Scientific Reports , volume=

    Super resolution DOA estimation based on deep neural network , author=. Scientific Reports , volume=. 2020 , publisher=

    2020

  25. [25]

    Generative-Adversarial-Network-Based Wireless Channel Modeling: Challenges and Opportunities , year=

    Yang, Yang and Li, Yang and Zhang, Wuxiong and Qin, Fei and Zhu, Pengcheng and Wang, Cheng-Xiang , journal=. Generative-Adversarial-Network-Based Wireless Channel Modeling: Challenges and Opportunities , year=

  26. [26]

    2019 IEEE International Symposium on Information Theory (ISIT) , pages=

    Solving linear inverse problems using generative models , author=. 2019 IEEE International Symposium on Information Theory (ISIT) , pages=. 2019 , organization=

    2019

  27. [27]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Photo-realistic single image super-resolution using a generative adversarial network , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  28. [28]

    European Conference on Computer Vision , pages=

    Srflow: Learning the super-resolution space with normalizing flow , author=. European Conference on Computer Vision , pages=. 2020 , organization=

    2020

  29. [29]

    Glow: Generative Flow with Invertible 1x1 Convolutions , url =

    Kingma, Durk P and Dhariwal, Prafulla , booktitle =. Glow: Generative Flow with Invertible 1x1 Convolutions , url =

  30. [30]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

    Normalizing flows: An introduction and review of current methods , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

  31. [31]

    Advances in Neural Information Processing Systems , volume=

    Neural spline flows , author=. Advances in Neural Information Processing Systems , volume=

  32. [32]

    Adam: A Method for Stochastic Optimization

    Adam: A method for stochastic optimization , author=. arXiv preprint arXiv:1412.6980 , year=

    work page internal anchor Pith review arXiv

  33. [33]

    Journal of Machine Learning Research , volume=

    Normalizing flows for probabilistic modeling and inference , author=. Journal of Machine Learning Research , volume=

  34. [34]

    International conference on machine learning , pages=

    Parallel wavenet: Fast high-fidelity speech synthesis , author=. International conference on machine learning , pages=. 2018 , organization=

    2018

  35. [35]

    International Conference on Machine Learning , pages=

    Neural autoregressive flows , author=. International Conference on Machine Learning , pages=. 2018 , organization=

    2018

  36. [36]

    3rd International Conference on Learning Representations,

    Laurent Dinh and David Krueger and Yoshua Bengio , editor =. 3rd International Conference on Learning Representations,. 2015 , url =

    2015

  37. [37]

    Electronic Journal of Statistics , volume=

    Estimating covariance and precision matrices along subspaces , author=. Electronic Journal of Statistics , volume=. 2021 , publisher=

    2021

  38. [38]

    Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages=

    A high-quality denoising dataset for smartphone cameras , author=. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages=

  39. [39]

    Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages=

    Image blind denoising with generative adversarial network based noise modeling , author=. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages=

  40. [40]

    IEEE Transactions on Image Processing , volume=

    Practical Poissonian-Gaussian noise modeling and fitting for single-image raw-data , author=. IEEE Transactions on Image Processing , volume=. 2008 , publisher=

    2008

  41. [41]

    IEEE Transactions on Image Processing , volume=

    Improved denoising via Poisson mixture modeling of image sensor noise , author=. IEEE Transactions on Image Processing , volume=. 2017 , publisher=

    2017

  42. [42]

    Advances in neural information processing systems , volume=

    Pytorch: An imperative style, high-performance deep learning library , author=. Advances in neural information processing systems , volume=

  43. [43]

    IEEE Communications Letters , volume=

    The Cramer-Rao bounds of hybrid TOA/RSS and TDOA/RSS location estimation schemes , author=. IEEE Communications Letters , volume=. 2004 , publisher=

    2004

  44. [44]

    2004 , publisher=

    Detection, estimation, and modulation theory, part I: detection, estimation, and linear modulation theory , author=. 2004 , publisher=

    2004

  45. [45]

    Information and the accuracy attainable in the estimation of statistical parameters , author=. Reson. J. Sci. Educ , volume=

  46. [46]

    IEEE Communications Magazine , volume=

    Generative-adversarial-network-based wireless channel modeling: Challenges and opportunities , author=. IEEE Communications Magazine , volume=. 2019 , publisher=

    2019

  47. [47]

    2021 IEEE 12th Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON) , pages=

    Neural Network Generative Models for Radio Frequency Data , author=. 2021 IEEE 12th Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON) , pages=. 2021 , organization=

    2021

  48. [48]

    International Conference on Artificial Intelligence and Statistics , pages=

    The expressive power of a class of normalizing flow models , author=. International Conference on Artificial Intelligence and Statistics , pages=. 2020 , organization=

    2020

  49. [49]

    GitHub repository , howpublished =

    Hai Victor Habi , title =. GitHub repository , howpublished =. 2022 , publisher =

    2022

  50. [50]

    Technical University of Denmark , volume=

    The matrix cookbook , author=. Technical University of Denmark , volume=

  51. [51]

    Density estimation using Real NVP

    Density estimation using real nvp , author=. arXiv preprint arXiv:1605.08803 , year=

    work page internal anchor Pith review arXiv

  52. [52]

    2006 , publisher=

    Theory of point estimation , author=. 2006 , publisher=

    2006

  53. [53]

    Advances in Neural Information Processing Systems , volume=

    Universal Approximation Using Well-Conditioned Normalizing Flows , author=. Advances in Neural Information Processing Systems , volume=

  54. [54]

    Gaussian Error Linear Units (GELUs)

    Gaussian error linear units (gelus) , author=. arXiv preprint arXiv:1606.08415 , year=

    work page internal anchor Pith review arXiv

  55. [55]

    The Annals of Statistics , volume=

    Robust multivariate mean estimation: the optimality of trimmed mean , author=. The Annals of Statistics , volume=. 2021 , publisher=

    2021

  56. [56]

    Electronic Journal of Statistics , volume=

    A general family of trimmed estimators for robust high-dimensional data analysis , author=. Electronic Journal of Statistics , volume=. 2018 , publisher=

    2018

  57. [57]

    and Poor, H.V

    Kassam, S.A. and Poor, H.V. , journal=. Robust techniques for signal processing: A survey , year=

  58. [58]

    arXiv preprint arXiv:2107.07232 , year=

    On the expressivity of bi-Lipschitz normalizing flows , author=. arXiv preprint arXiv:2107.07232 , year=

    work page arXiv

  59. [59]

    Economic Geology , volume =

    Principles of geostatistics , author =. Economic Geology , volume =. 1963 , doi =

    1963

  60. [60]

    2005 , month = mar, url =

    2005

  61. [61]

    ICML Workshop on Invertible Neural Networks, Normalizing Flows, and Explicit Likelihood Models , year=

    Universal Approximation of Residual Flows in Maximum Mean Discrepancy , author=. ICML Workshop on Invertible Neural Networks, Normalizing Flows, and Explicit Likelihood Models , year=

  62. [62]

    IEEE Transactions on information theory , volume=

    Single tone parameter estimation from discrete-time observations , author=. IEEE Transactions on information theory , volume=. 1974 , publisher=

    1974

  63. [63]

    IEEE Transactions on Signal Processing , volume=

    Effects of sampling and quantization on single-tone frequency estimation , author=. IEEE Transactions on Signal Processing , volume=. 2000 , publisher=

    2000

  64. [64]

    IEEE transactions on wireless communications , volume=

    On the Cramer-Rao bound for carrier frequency estimation in the presence of phase noise , author=. IEEE transactions on wireless communications , volume=. 2007 , publisher=

    2007

  65. [65]

    IEEE Transactions on Signal Processing , volume=

    Joint ML/MAP Estimation of the Frequency and Phase of a Single Sinusoid With Wiener Carrier Phase Noise , author=. IEEE Transactions on Signal Processing , volume=. 2021 , publisher=

    2021

  66. [66]

    Advances in Neural Information Processing Systems , volume=

    Data-driven estimation of sinusoid frequencies , author=. Advances in Neural Information Processing Systems , volume=

  67. [67]

    Energies , volume=

    Estimation of modal parameters for inter-area oscillations analysis by a machine learning approach with offline training , author=. Energies , volume=. 2020 , publisher=

    2020

  68. [68]

    arXiv preprint arXiv:2203.03695 , year=

    Learning to Bound: A Generative Cram 'er-Rao Bound , author=. arXiv preprint arXiv:2203.03695 , year=

    work page arXiv

  69. [69]

    International Conference on Machine Learning , pages=

    Flow++: Improving flow-based generative models with variational dequantization and architecture design , author=. International Conference on Machine Learning , pages=. 2019 , organization=

    2019

  70. [70]

    A note on the evaluation of generative models

    A note on the evaluation of generative models , author=. arXiv preprint arXiv:1511.01844 , year=

    work page Pith review arXiv

  71. [71]

    Conference on Learning Theory , pages=

    Non-convex learning via stochastic gradient langevin dynamics: a nonasymptotic analysis , author=. Conference on Learning Theory , pages=. 2017 , organization=

    2017

  72. [72]

    Advances in Neural Information Processing Systems , volume=

    Global convergence of Langevin dynamics based algorithms for nonconvex optimization , author=. Advances in Neural Information Processing Systems , volume=

  73. [73]

    Estimation of Non-Normalized Statistical Models by Score Matching , year =

    Hyv\". Estimation of Non-Normalized Statistical Models by Score Matching , year =. J. Mach. Learn. Res. , month = dec, pages =

  74. [74]

    arXiv preprint arXiv:2304.04740 , year=

    Reflected Diffusion Models , author=. arXiv preprint arXiv:2304.04740 , year=

    work page arXiv

  75. [75]

    arXiv preprint arXiv:2203.10215 , year=

    Convergence error analysis of reflected gradient langevin dynamics for globally optimizing non-convex constrained problems , author=. arXiv preprint arXiv:2203.10215 , year=

    work page arXiv

  76. [76]

    Journal of Machine Learning Research , volume=

    Estimating density models with truncation boundaries using score matching , author=. Journal of Machine Learning Research , volume=

  77. [77]

    Journal of statistical computation and simulation , volume=

    Stochastic gradient Langevin dynamics with adaptive drifts , author=. Journal of statistical computation and simulation , volume=. 2022 , publisher=

    2022

  78. [78]

    Computational statistics & data analysis , volume=

    Some extensions of score matching , author=. Computational statistics & data analysis , volume=. 2007 , publisher=

    2007

  79. [79]

    The Journal of Machine Learning Research , volume=

    Generalized score matching for non-negative data , author=. The Journal of Machine Learning Research , volume=. 2019 , publisher=

    2019

  80. [80]

    Information and Inference: A Journal of the IMA , volume=

    Generalized score matching for general domains , author=. Information and Inference: A Journal of the IMA , volume=. 2022 , publisher=

    2022

Showing first 80 references.