Physics-Constrained Synthetic Training for Sub-Terahertz Channel Rainfall Sensing

Hong Ling; Jiabiao Zhao; Jianjun Ma; Jie Yang; Kaixin Sun; Kefeng Huang; Mingxia Zhang; Wanzhu Chang; Yuheng Song

arxiv: 2606.21857 · v1 · pith:XOWC2SZWnew · submitted 2026-06-20 · ⚛️ physics.app-ph

Physics-Constrained Synthetic Training for Sub-Terahertz Channel Rainfall Sensing

Wanzhu Chang , Yuheng Song , Jiabiao Zhao , Mingxia Zhang , Jie Yang , Kefeng Huang , Kaixin Sun , Hong Ling

show 1 more author

Jianjun Ma

This is my paper

Pith reviewed 2026-06-26 11:20 UTC · model grok-4.3

classification ⚛️ physics.app-ph

keywords terahertz channelsrainfall sensingsynthetic trainingraindrop size distributionphysics-constrained learningsub-terahertzchannel modelingzero-shot estimation

0 comments

The pith

Synthetic data generated from rain attenuation models and multiple drop-size priors trains models that estimate rainfall from sub-THz received-power sequences without any real labeled examples.

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

The paper establishes that physics-constrained synthetic training data, built from ITU-R P.838-3 attenuation formulas, Mie scattering, and four different raindrop-size-distribution priors plus controlled fluctuations, can replace scarce measured datasets for training a rainfall estimator. A hybrid attention-convolution network called RainFormer is trained on short sequences of synthetic received power and then applied directly to real 140 GHz and 229 GHz measurements from a 41.5 m outdoor link. The resulting estimates remain rank-consistent and physically plausible across the different priors, while the spread among those priors supplies a bound on uncertainty caused by the unknown drop-size distribution along the actual path.

Core claim

DSD-bracketed synthetic training produces zero-shot models whose rainfall-rate estimates on measured THz channels are rank-consistent and physically interpretable, with the spread across the four DSD priors bounding the uncertainty that arises from the unobservable path drop-size distribution.

What carries the argument

DSD-bracketed synthetic training that combines ITU-R P.838-3 and Mie-theory attenuation with four raindrop-size-distribution priors to generate labeled sequences whose inter-prior variation brackets path DSD uncertainty.

If this is right

Ablation experiments show that explicit physical-statistical descriptors and preservation of temporal order supply most of the predictive signal.
Convolution and attention layers act as complementary refinements rather than primary drivers.
The approach supplies a practical route to rainfall sensing at sub-THz frequencies when operational labeled datasets do not exist.
Inter-prior spread offers a built-in indicator of uncertainty attributable to unknown path drop-size distribution.

Where Pith is reading between the lines

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

The same bracketing strategy could be tested on longer links or different carrier frequencies once calibration data become available.
Operational systems could run the model on each prior in parallel to report both a point estimate and an explicit uncertainty interval.
If the spread reliably bounds error, the method might transfer to other remote-sensing tasks whose key microphysical parameters remain unobserved.

Load-bearing premise

The statistical properties observed in the 41.5 m outdoor campaign at 140 GHz and 229 GHz are representative of real channel behavior under rainfall, allowing the synthetic data to generalize to unseen measured events.

What would settle it

An experiment that measures both received power and independent rain-rate plus drop-size-distribution data along the same path, then checks whether the inter-prior spread of model outputs actually contains the error relative to the measured rain rate.

read the original abstract

Terahertz channels can serve as opportunistic rainfall sensors because rain-induced extinction couples received power to rainfall intensity. Unlike satellite, radar, and commercial microwave link retrieval, THz channel rainfall estimation lacks large operational datasets that supervised learning requires. This article uses an outdoor campaign over a 41.5 m THz channel at 140 GHz and 229 GHz to calibrate the channel statistical properties, then synthesizing physics constrained training data that combine the ITU-R P.838-3 and Mie-theory rain attenuation across four raindrop-size-distribution priors with controlled stochastic fluctuations. RainFormer, a hybrid attention-convolution network, maps a short received-power sequence to rainfall rate by fusing local fluctuation structure, long-range temporal dependencies, and explicit physical-statistical descriptors. Ablation shows that the explicit descriptors and temporal-order information carry most of the predictive information, with convolution and attention acting as complementary refinements. Applied zero-shot to measured rainfall, the synthetic-trained models produce rank-consistent, physically interpretable estimates whose inter-prior spread bounds the uncertainty arising from the unobservable path DSD (raindrop size distribution), establishing DSD-bracketed synthetic training as a viable foundation for THz rainfall sensing under severe data scarcity.

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.

Desk Editor's Note private letter to a colleague

They calibrate a short THz link to build synthetic training data from standard attenuation models plus four DSD priors, then train RainFormer and get rank-consistent zero-shot estimates on real measurements.

read the letter

The main contribution is the end-to-end pipeline: a 41.5 m outdoor campaign at 140 and 229 GHz supplies the channel statistics, which are then used to synthesize sequences via ITU-R P.838-3 and Mie theory across four raindrop-size priors with added stochastic fluctuations. RainFormer fuses local structure, longer temporal context, and explicit physical descriptors to map received-power sequences to rainfall rate. Ablations indicate the physical descriptors and temporal ordering do most of the work, and the zero-shot results on measured rain are at least rank-consistent with the inter-prior spread offered as a bracket on DSD uncertainty.

That approach is a reasonable incremental step for data-scarce THz sensing. It stays grounded in established physics models rather than fitting directly to the target data, and the multi-prior bracketing is a clean way to surface one source of uncertainty.

The clearest limitation is the calibration step itself. A 41.5 m path, even with an outdoor campaign, is short and likely limited in rain-type diversity and duration; it is not obvious that the extracted fluctuation statistics and attenuation behavior will match longer paths or different meteorological conditions. If they do not, both the generalization and the uncertainty-bounding interpretation rest on weaker ground. The abstract reports ablations and zero-shot performance but gives no quantitative metrics or error analysis, so the practical accuracy remains hard to judge from what is shown.

This is for groups working on sub-THz links or opportunistic weather sensing. Readers who need a concrete example of physics-constrained synthetic training will find the setup useful. The work is coherent on its own terms and grounded enough to deserve a serious referee, though the review will need the full numbers and more detail on how representative the campaign statistics actually are.

Referee Report

2 major / 1 minor

Summary. The manuscript claims that an outdoor 41.5 m THz channel campaign at 140 GHz and 229 GHz can be used to calibrate statistical properties, enabling generation of physics-constrained synthetic training data via ITU-R P.838-3, Mie theory, and four DSD priors with controlled stochastic fluctuations. A hybrid RainFormer network trained on this data performs zero-shot rainfall estimation on measured sequences, yielding rank-consistent, physically interpretable outputs whose inter-prior spread bounds unobservable path DSD uncertainty, thereby establishing DSD-bracketed synthetic training as viable under severe data scarcity. Ablations indicate that explicit physical-statistical descriptors and temporal order carry most predictive power.

Significance. If the central claim holds, the work provides a concrete route to supervised learning for sub-THz opportunistic rainfall sensing without requiring large operational datasets, by anchoring synthetic traces in standard physical models plus limited calibration. The demonstration that inter-prior spread can bracket DSD-induced uncertainty offers a falsifiable uncertainty quantification approach that could generalize to other link-based environmental sensing problems where direct labeled data are scarce.

major comments (2)

[Abstract] Abstract and results description: the claim that synthetic-trained models produce 'rank-consistent' zero-shot estimates on measured rainfall whose inter-prior spread 'bounds the uncertainty arising from the unobservable path DSD' is load-bearing for the central contribution, yet no quantitative metrics (RMSE, Spearman rank correlation, or coverage of measured rates by the prior envelope) or error analysis are supplied to support it.
[Calibration and synthesis procedure] Outdoor campaign and data-generation section: the assumption that fluctuation statistics and attenuation properties extracted from the 41.5 m, 140/229 GHz campaign are representative of real-channel behavior under varying rainfall (different DSD regimes, scintillation, longer paths) is required for generalization; no cross-validation against independent rain events, longer-path data, or sensitivity tests to campaign duration/rain-type diversity is reported.

minor comments (1)

Notation for the four DSD priors and the explicit physical-statistical descriptors should be defined with symbols and units in a dedicated table or subsection for reproducibility.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the constructive and detailed review. We address each major comment below, indicating planned revisions where the manuscript can be strengthened.

read point-by-point responses

Referee: [Abstract] Abstract and results description: the claim that synthetic-trained models produce 'rank-consistent' zero-shot estimates on measured rainfall whose inter-prior spread 'bounds the uncertainty arising from the unobservable path DSD' is load-bearing for the central contribution, yet no quantitative metrics (RMSE, Spearman rank correlation, or coverage of measured rates by the prior envelope) or error analysis are supplied to support it.

Authors: We agree that explicit quantitative metrics are required to substantiate the central claims. In the revised manuscript we will add RMSE, Spearman rank correlation, an assessment of coverage of measured rates by the inter-prior envelope, and accompanying error analysis to the results section and abstract. revision: yes
Referee: [Calibration and synthesis procedure] Outdoor campaign and data-generation section: the assumption that fluctuation statistics and attenuation properties extracted from the 41.5 m, 140/229 GHz campaign are representative of real-channel behavior under varying rainfall (different DSD regimes, scintillation, longer paths) is required for generalization; no cross-validation against independent rain events, longer-path data, or sensitivity tests to campaign duration/rain-type diversity is reported.

Authors: The 41.5 m campaign supplies the statistical calibration, while ITU-R P.838-3 and Mie theory supply the physical basis intended to support generalization across DSD regimes. We will add sensitivity tests to campaign duration and rain-type assumptions in the revision. Full cross-validation against independent rain events or longer paths cannot be performed with the existing dataset. revision: partial

standing simulated objections not resolved

Cross-validation against independent rain events and longer-path data, which would require additional experimental campaigns outside the scope of the present work.

Circularity Check

0 steps flagged

No circularity; derivation uses independent calibration and external models

full rationale

The paper calibrates statistical properties from a separate 41.5 m outdoor measurement campaign, then generates synthetic training data via external ITU-R P.838-3 and Mie theory plus four DSD priors. The RainFormer is trained exclusively on this synthetic data and evaluated zero-shot on held-out measured rainfall. No equation or claim reduces a reported prediction to a quantity fitted directly on the target rainfall data, and no load-bearing step relies on self-citation. The inter-prior spread is an explicit bracketing device rather than a fitted output. The chain is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 3 axioms · 0 invented entities

The central claim rests on standard rain attenuation models and the representativeness of a short-path calibration campaign; no new entities are postulated.

free parameters (1)

stochastic fluctuation parameters
Controlled random variations added to attenuation models during data synthesis; their exact tuning is not detailed in the abstract.

axioms (3)

domain assumption ITU-R P.838-3 provides accurate specific attenuation for 140 GHz and 229 GHz rain
Directly invoked to generate synthetic training data
domain assumption Mie scattering theory correctly describes raindrop extinction at these frequencies
Combined with ITU-R model for the synthetic data
domain assumption The 41.5 m campaign statistics are representative of path behavior under rainfall
Used to constrain the synthetic data generation

pith-pipeline@v0.9.1-grok · 5763 in / 1479 out tokens · 33161 ms · 2026-06-26T11:20:41.743952+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

52 extracted references

[1]

6G: The Next Horizon: From Connected People and Things to Connected Intelligence,

B. Rong, "6G: The Next Horizon: From Connected People and Things to Connected Intelligence," IEEE Wireless Communications, vol. 28, no. 5, pp. 8-8, 2021

2021
[2]

Integrated Sensing and Communications: Toward Dual-Functional Wireless Networks for 6G and Beyond,

F. Liu et al., "Integrated Sensing and Communications: Toward Dual-Functional Wireless Networks for 6G and Beyond," IEEE Journal on Selected Areas in Communications, vol. 40, no. 6, pp. 1728-1767, 2022

2022
[3]

Terahertz Communications and Sensing for 6G and Beyond: A Comprehensive Review,

W. Jiang et al., "Terahertz Communications and Sensing for 6G and Beyond: A Comprehensive Review," IEEE Communications Surveys & Tutorials, vol. 26, no. 4, pp. 2326-2381, 2024

2024
[4]

Terahertz-Band Integrated Sensing and Communications: Challenges and Opportunities,

A. M. Elbir, K. V. Mishra, S. Chatzinotas, and M. Bennis, "Terahertz-Band Integrated Sensing and Communications: Challenges and Opportunities," IEEE Aerospace and Electronic Systems Magazine, vol. 39, no. 12, pp. 38-49, 2024

2024
[5]

Rain Attenuation at Millimeter Wave and Low-THz Frequencies,

F. Norouzian et al., "Rain Attenuation at Millimeter Wave and Low-THz Frequencies," IEEE Transactions on Antennas and Propagation, vol. 68, no. 1, pp. 421-431, 2020

2020
[6]

Terahertz channel power and BER performance in rain,

Y. Song et al., "Terahertz channel power and BER performance in rain," Opt Express, vol. 33, no. 5, pp. 11336-11349, Mar 10 2025

2025
[7]

Effect of Rain Attenuation for a 10-Gb/s 120-GHz-Band Millimeter-Wave Wireless Link,

A. Hirata et al., "Effect of Rain Attenuation for a 10-Gb/s 120-GHz-Band Millimeter-Wave Wireless Link," IEEE Transactions on Microwave Theory and Techniques, vol. 57, no. 12, pp. 3099-3105, 2009

2009
[8]

Backscatter and attenuation by falling snow and rain at 96, 140, and 225 GHz,

J. Nemarich, R. J. Wellman, and J. Lacombe, "Backscatter and attenuation by falling snow and rain at 96, 140, and 225 GHz," IEEE Transactions on Geoscience and Remote Sensing, vol. 26, no. 3, pp. 319-329, 1988

1988
[9]

Rain Attenuation in the Microwave-to-Terahertz Waveband,

S. Ishii, M. Kinugawa, S. Wakiyama, S. Sayama, and T. Kamei, "Rain Attenuation in the Microwave-to-Terahertz Waveband," Wireless Engineering and Technology, vol. 07, no. 02, pp. 59-66, 2016

2016
[10]

Review of weather impact on outdoor terahertz wireless communication links,

J. F. Federici, J. Ma, and L. Moeller, "Review of weather impact on outdoor terahertz wireless communication links," Nano Communication Networks, 10, pp. 12-26, 2016

2016
[11]

Measurement and Modeling on Terahertz Channels in Rain,

P. Li et al., "Measurement and Modeling on Terahertz Channels in Rain," Journal of Infrared, Millimeter, and Terahertz Waves, vol. 46, no. 1, 2024

2024
[12]

The Global Precipitation Measurement (Gpm) Mission for Science and Society,

G. Skofronick-Jackson et al., "The Global Precipitation Measurement (Gpm) Mission for Science and Society," Bull Am Meteorol Soc, vol. 98, no. 8, pp. 1679-1695, Aug 2017

2017
[13]

Integrated Multi-satellite Retrievals for the Global Precipitation Measurement (GPM) Mission (IMERG),

G. J. Huffman et al., "Integrated Multi-satellite Retrievals for the Global Precipitation Measurement (GPM) Mission (IMERG)," in Satellite Precipitation Measurement, (Advances in Global Change Research, 2020, ch. Chapter 19, pp. 343-353

2020
[14]

CMORPH: A Method that Produces Global Precipitation Estimates from Passive Microwave and Infrared Data at High Spatial and Temporal Resolution,

R. J. Joyce, J. E. Janowiak, P. A. Arkin, and P. Xie, "CMORPH: A Method that Produces Global Precipitation Estimates from Passive Microwave and Infrared Data at High Spatial and Temporal Resolution," Journal of Hydrometeorology, vol. 5, no. 3, pp. 487-503, 2004

2004
[15]

Global Satellite Mapping of Precipitation (GSMaP) Products in the GPM Era,

T. Kubota et al., "Global Satellite Mapping of Precipitation (GSMaP) Products in the GPM Era," in Satellite Precipitation Measurement, (Advances in Global Change Research, 2020, ch. Chapter 20, pp. 355-373

2020
[16]

PERSIANN-CDR: Daily Precipitation Climate Data Record from Multisatellite Observations for Hydrological and Climate Studies,

D. K. Braithwaite et al., "PERSIANN-CDR: Daily Precipitation Climate Data Record from Multisatellite Observations for Hydrological and Climate Studies," Bulletin of the American Meteorological Society, vol. 96, no. 1, pp. 69-83, 2015

2015
[17]

Deep Neural Network High Spatiotemporal Resolution Precipitation Estimation (Deep-STEP) Using Passive Microwave and Infrared Data,

V. A. Gorooh, A. A. Asanjan, P. Nguyen, K. Hsu, and S. Sorooshian, "Deep Neural Network High Spatiotemporal Resolution Precipitation Estimation (Deep-STEP) Using Passive Microwave and Infrared Data," Journal of Hydrometeorology, vol. 23, no. 4, pp. 597-617, 2022

2022
[18]

PrecipGAN: Merging Microwave and Infrared Data for Satellite Precipitation Estimation Using Generative Adversarial Network,

C. Wang, G. Tang, and P. Gentine, "PrecipGAN: Merging Microwave and Infrared Data for Satellite Precipitation Estimation Using Generative Adversarial Network," Geophysical Research Letters, vol. 48, no. 5, 2021

2021
[19]

Global Precipitation Nowcasting of Integrated Multi-satellitE Retrievals for GPM: A U-Net Convolutional LSTM Architecture,

R. Rahimi, P. Ravirathinam, A. Ebtehaj, A. Behrangi, J. Tan, and V. Kumar, "Global Precipitation Nowcasting of Integrated Multi-satellitE Retrievals for GPM: A U-Net Convolutional LSTM Architecture," Journal of Hydrometeorology, vol. 25, no. 6, pp. 947-963, 2024

2024
[20]

RainNet v1.0: a convolutional neural network for radar-based precipitation nowcasting,

G. Ayzel, T. Scheffer, and M. Heistermann, "RainNet v1.0: a convolutional neural network for radar-based precipitation nowcasting," Geoscientific Model Development, vol. 13, no. 6, pp. 2631-2644, 2020

2020
[21]

Precipitation nowcasting using transformer-based generative models and transfer learning for improved disaster preparedness,

M. J. Piran, X. Wang, H. J. Kim, and H. H. Kwon, "Precipitation nowcasting using transformer-based generative models and transfer learning for improved disaster preparedness," International Journal of Applied Earth Observation and Geoinformation, vol. 132, 2024

2024
[22]

Environmental monitoring by wireless communication networks,

H. Messer, A. Zinevich, and P. Alpert, "Environmental monitoring by wireless communication networks," Science, vol. 312, no. 5774, p. 713, 2006

2006
[23]

Rainfall measurement using radio links from cellular communication networks,

H. Leijnse, R. Uijlenhoet, and J. N. M. Stricker, "Rainfall measurement using radio links from cellular communication networks," Water Resources Research, vol. 43, no. 3, 2007

2007
[24]

Rain event detection in commercial microwave link attenuation data using convolutional neural networks,

J. Polz, C. Chwala, M. Graf, and H. Kunstmann, "Rain event detection in commercial microwave link attenuation data using convolutional neural networks," Atmospheric Measurement Techniques, vol. 13, no. 7, pp. 3835-3853, 2020

2020
[25]

Commercial microwave link networks for rainfall observation: Assessment of the current status and future challenges,

C. Chwala and H. Kunstmann, "Commercial microwave link networks for rainfall observation: Assessment of the current status and future challenges," WIREs Water, vol. 6, no. 2, 2019

2019
[26]

A Review on Rainfall Measurement Based on Commercial Microwave Links in Wireless Cellular Networks,

B. Lian, Z. Wei, X. Sun, Z. Li, and J. Zhao, "A Review on Rainfall Measurement Based on Commercial Microwave Links in Wireless Cellular Networks," Sensors (Basel), vol. 22, no. 12, Jun 10 2022

2022
[27]

Rain Attenuation Statistics for a 120-GHz-Band Wireless Link,

A. Hirata, H. Takahashi, T. Kosugi, K. Murata, K. Naoya, and Y. Kado, "Rain Attenuation Statistics for a 120-GHz-Band Wireless Link," Ieee Mtt S Int Micr, pp. 933, 2009

2009
[28]

Specific attenuation model for rain for use in prediction methods,

ITU-R, "Specific attenuation model for rain for use in prediction methods," Geneva, Switzerland, Recommendation ITU-R P.838-3, 2005

2005
[29]

Statistical theory of the Marshall-Palmer distribution of raindrops,

L. Yangang, "Statistical theory of the Marshall-Palmer distribution of raindrops," Atmospheric Environment. Part A. General Topics, vol. 27, no. 1, pp. 15-19, 1993

1993
[30]

Feloni, K

E. Feloni, K. Kotsifakis, N. Dervos, G. Giavis, and E. Baltas, Analysis of Joss-Waldvogel disdrometer measurements in rainfall events (Fifth International Conference on Remote Sensing and Geoinformation of the Environment (RSCy2017)). SPIE, 2017

2017
[31]

Drop-Size Distributions and Mie Computations for Rain,

C. Mätzler, "Drop-Size Distributions and Mie Computations for Rain," Institut für Angewandte Physik, Universität Bern, Bern, Switzerland, November 2002 2002

2002
[32]

Microwave link dual-wavelength measurements of path-average attenuation for the estimation of drop size distributions and rainfall,

R. F. Rincon and R. H. Lang, "Microwave link dual-wavelength measurements of path-average attenuation for the estimation of drop size distributions and rainfall," IEEE Transactions on Geoscience and Remote Sensing, vol. 40, no. 4, pp. 760-770, 2002

2002
[33]

Estimating raindrop size distributions using microwave link measurements: potential and limitations,

T. C. van Leth, H. Leijnse, A. Overeem, and R. Uijlenhoet, "Estimating raindrop size distributions using microwave link measurements: potential and limitations," Atmospheric Measurement Techniques, vol. 13, no. 4, pp. 1797-1815, 2020

2020
[34]

Measurements and models for 38-GHz point-to-multipoint radiowave propagation,

X. Hao, T. S. Rappaport, R. J. Boyle, and J. H. Schaffner, "Measurements and models for 38-GHz point-to-multipoint radiowave propagation," IEEE Journal on Selected Areas in Communications, vol. 18, no. 3, pp. 310-321, 2000

2000
[35]

Bootstrap Methods: Another Look at the Jackknife,

B. Efron, "Bootstrap Methods: Another Look at the Jackknife," The Annals of Statistics, vol. 7, no. 1, 1979

1979
[36]

Time- and Frequency-Varying K-Factor of Non-Stationary Vehicular Channels for Safety-Relevant Scenarios,

L. Bernado, T. Zemen, F. Tufvesson, A. F. Molisch, and C. F. Mecklenbrauker, "Time- and Frequency-Varying K-Factor of Non-Stationary Vehicular Channels for Safety-Relevant Scenarios," IEEE Transactions on Intelligent Transportation Systems, vol. 16, no. 2, pp. 1007-1017, 2015

2015
[37]

On the estimation of the K parameter for the Rice fading distribution,

A. Abdi, C. Tepedelenlioglu, M. Kaveh, and G. Giannakis, "On the estimation of the K parameter for the Rice fading distribution," IEEE Communications Letters, vol. 5, no. 3, pp. 92-94, 2001

2001
[38]

Rain: relaxations in the sky,

O. Peters and K. Christensen, "Rain: relaxations in the sky," Phys Rev E Stat Nonlin Soft Matter Phys, vol. 66, no. 3 Pt 2A, p. 036120, Sep 2002

2002
[39]

Understanding and improving layer normalization

J. Xu, et al. "Understanding and improving layer normalization." presented at the Advances in neural information processing systems 32, 2019

2019
[40]

Attention is All you Need,

A. Vaswani et al., "Attention is All you Need," presented at the Advances in Neural Information Processing Systems 30, 2017

2017
[41]

Language Modeling with Gated Convolutional Networks,

Y. N. Dauphin, A. Fan, M. Auli, and D. Grangier, "Language Modeling with Gated Convolutional Networks," presented at the Proceedings of the 34th International Conference on Machine Learning, 2017

2017
[42]

Dropout: A Simple Way to Prevent Neural Networks from Overfitting,

N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, "Dropout: A Simple Way to Prevent Neural Networks from Overfitting," (in English), J Mach Learn Res, vol. 15, pp. 1929-1958, Jun 2014

1929
[43]

A time series forest for classification and feature extraction,

H. Deng, G. Runger, E. Tuv, and M. Vladimir, "A time series forest for classification and feature extraction," Information Sciences, vol. 239, pp. 142-153, 2013

2013
[44]

Domain randomization for transferring deep neural networks from simulation to the real world,

J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel, "Domain randomization for transferring deep neural networks from simulation to the real world," presented at the 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017

2017
[45]

Physics-informed machine learning,

G. E. Karniadakis, I. G. Kevrekidis, L. Lu, P. Perdikaris, S. Wang, and L. Yang, "Physics-informed machine learning," Nature Reviews Physics, vol. 3, no. 6, pp. 422-440, 2021

2021
[46]

Are Transformers Effective for Time Series Forecasting?,

A. Zeng, M. Chen, L. Zhang, and Q. Xu, "Are Transformers Effective for Time Series Forecasting?," Proceedings of the AAAI Conference on Artificial Intelligence, vol. 37, no. 9, pp. 11121-11128, 2023

2023
[47]

Radar Reflectivity and Meteorological Factors Merging‐Based Precipitation Estimation Neural Network,

Y. Zhang, S. Chen, W. Tian, and S. Chen, "Radar Reflectivity and Meteorological Factors Merging‐Based Precipitation Estimation Neural Network," Earth and Space Science, vol. 8, no. 10, 2021

2021
[48]

Deep learning for an improved prediction of rainfall retrievals from commercial microwave links

J. Pudashine, et al., "Deep learning for an improved prediction of rainfall retrievals from commercial microwave links." Water Resources Research 56, no. 7, pp. e2019WR026255, 2020

2020
[49]

Transformers in Time Series: A Survey,

Q. Wen et al., "Transformers in Time Series: A Survey," presented at the Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, 2023

2023
[50]

A Time Series Is Worth 64 Words: Long-term Forecasting with Transformers,

Y. Nie, N. H. Nguyen, P. Sinthong, and J. Kalagnanam, "A Time Series Is Worth 64 Words: Long-term Forecasting with Transformers," presented at the International Conference on Learning Representations, 2023

2023
[51]

Root mean square error (RMSE) or mean absolute error (MAE)? – Arguments against avoiding RMSE in the literature,

T. Chai and R. R. Draxler, "Root mean square error (RMSE) or mean absolute error (MAE)? – Arguments against avoiding RMSE in the literature," Geoscientific Model Development, vol. 7, no. 3, pp. 1247-1250, 2014

2014
[52]

Conformer: Convolution-augmented Transformer for Speech Recognition,

A. Gulati et al., "Conformer: Convolution-augmented Transformer for Speech Recognition," presented at the Interspeech 2020, 2020

2020

[1] [1]

6G: The Next Horizon: From Connected People and Things to Connected Intelligence,

B. Rong, "6G: The Next Horizon: From Connected People and Things to Connected Intelligence," IEEE Wireless Communications, vol. 28, no. 5, pp. 8-8, 2021

2021

[2] [2]

Integrated Sensing and Communications: Toward Dual-Functional Wireless Networks for 6G and Beyond,

F. Liu et al., "Integrated Sensing and Communications: Toward Dual-Functional Wireless Networks for 6G and Beyond," IEEE Journal on Selected Areas in Communications, vol. 40, no. 6, pp. 1728-1767, 2022

2022

[3] [3]

Terahertz Communications and Sensing for 6G and Beyond: A Comprehensive Review,

W. Jiang et al., "Terahertz Communications and Sensing for 6G and Beyond: A Comprehensive Review," IEEE Communications Surveys & Tutorials, vol. 26, no. 4, pp. 2326-2381, 2024

2024

[4] [4]

Terahertz-Band Integrated Sensing and Communications: Challenges and Opportunities,

A. M. Elbir, K. V. Mishra, S. Chatzinotas, and M. Bennis, "Terahertz-Band Integrated Sensing and Communications: Challenges and Opportunities," IEEE Aerospace and Electronic Systems Magazine, vol. 39, no. 12, pp. 38-49, 2024

2024

[5] [5]

Rain Attenuation at Millimeter Wave and Low-THz Frequencies,

F. Norouzian et al., "Rain Attenuation at Millimeter Wave and Low-THz Frequencies," IEEE Transactions on Antennas and Propagation, vol. 68, no. 1, pp. 421-431, 2020

2020

[6] [6]

Terahertz channel power and BER performance in rain,

Y. Song et al., "Terahertz channel power and BER performance in rain," Opt Express, vol. 33, no. 5, pp. 11336-11349, Mar 10 2025

2025

[7] [7]

Effect of Rain Attenuation for a 10-Gb/s 120-GHz-Band Millimeter-Wave Wireless Link,

A. Hirata et al., "Effect of Rain Attenuation for a 10-Gb/s 120-GHz-Band Millimeter-Wave Wireless Link," IEEE Transactions on Microwave Theory and Techniques, vol. 57, no. 12, pp. 3099-3105, 2009

2009

[8] [8]

Backscatter and attenuation by falling snow and rain at 96, 140, and 225 GHz,

J. Nemarich, R. J. Wellman, and J. Lacombe, "Backscatter and attenuation by falling snow and rain at 96, 140, and 225 GHz," IEEE Transactions on Geoscience and Remote Sensing, vol. 26, no. 3, pp. 319-329, 1988

1988

[9] [9]

Rain Attenuation in the Microwave-to-Terahertz Waveband,

S. Ishii, M. Kinugawa, S. Wakiyama, S. Sayama, and T. Kamei, "Rain Attenuation in the Microwave-to-Terahertz Waveband," Wireless Engineering and Technology, vol. 07, no. 02, pp. 59-66, 2016

2016

[10] [10]

Review of weather impact on outdoor terahertz wireless communication links,

J. F. Federici, J. Ma, and L. Moeller, "Review of weather impact on outdoor terahertz wireless communication links," Nano Communication Networks, 10, pp. 12-26, 2016

2016

[11] [11]

Measurement and Modeling on Terahertz Channels in Rain,

P. Li et al., "Measurement and Modeling on Terahertz Channels in Rain," Journal of Infrared, Millimeter, and Terahertz Waves, vol. 46, no. 1, 2024

2024

[12] [12]

The Global Precipitation Measurement (Gpm) Mission for Science and Society,

G. Skofronick-Jackson et al., "The Global Precipitation Measurement (Gpm) Mission for Science and Society," Bull Am Meteorol Soc, vol. 98, no. 8, pp. 1679-1695, Aug 2017

2017

[13] [13]

Integrated Multi-satellite Retrievals for the Global Precipitation Measurement (GPM) Mission (IMERG),

G. J. Huffman et al., "Integrated Multi-satellite Retrievals for the Global Precipitation Measurement (GPM) Mission (IMERG)," in Satellite Precipitation Measurement, (Advances in Global Change Research, 2020, ch. Chapter 19, pp. 343-353

2020

[14] [14]

CMORPH: A Method that Produces Global Precipitation Estimates from Passive Microwave and Infrared Data at High Spatial and Temporal Resolution,

R. J. Joyce, J. E. Janowiak, P. A. Arkin, and P. Xie, "CMORPH: A Method that Produces Global Precipitation Estimates from Passive Microwave and Infrared Data at High Spatial and Temporal Resolution," Journal of Hydrometeorology, vol. 5, no. 3, pp. 487-503, 2004

2004

[15] [15]

Global Satellite Mapping of Precipitation (GSMaP) Products in the GPM Era,

T. Kubota et al., "Global Satellite Mapping of Precipitation (GSMaP) Products in the GPM Era," in Satellite Precipitation Measurement, (Advances in Global Change Research, 2020, ch. Chapter 20, pp. 355-373

2020

[16] [16]

PERSIANN-CDR: Daily Precipitation Climate Data Record from Multisatellite Observations for Hydrological and Climate Studies,

D. K. Braithwaite et al., "PERSIANN-CDR: Daily Precipitation Climate Data Record from Multisatellite Observations for Hydrological and Climate Studies," Bulletin of the American Meteorological Society, vol. 96, no. 1, pp. 69-83, 2015

2015

[17] [17]

Deep Neural Network High Spatiotemporal Resolution Precipitation Estimation (Deep-STEP) Using Passive Microwave and Infrared Data,

V. A. Gorooh, A. A. Asanjan, P. Nguyen, K. Hsu, and S. Sorooshian, "Deep Neural Network High Spatiotemporal Resolution Precipitation Estimation (Deep-STEP) Using Passive Microwave and Infrared Data," Journal of Hydrometeorology, vol. 23, no. 4, pp. 597-617, 2022

2022

[18] [18]

PrecipGAN: Merging Microwave and Infrared Data for Satellite Precipitation Estimation Using Generative Adversarial Network,

C. Wang, G. Tang, and P. Gentine, "PrecipGAN: Merging Microwave and Infrared Data for Satellite Precipitation Estimation Using Generative Adversarial Network," Geophysical Research Letters, vol. 48, no. 5, 2021

2021

[19] [19]

Global Precipitation Nowcasting of Integrated Multi-satellitE Retrievals for GPM: A U-Net Convolutional LSTM Architecture,

R. Rahimi, P. Ravirathinam, A. Ebtehaj, A. Behrangi, J. Tan, and V. Kumar, "Global Precipitation Nowcasting of Integrated Multi-satellitE Retrievals for GPM: A U-Net Convolutional LSTM Architecture," Journal of Hydrometeorology, vol. 25, no. 6, pp. 947-963, 2024

2024

[20] [20]

RainNet v1.0: a convolutional neural network for radar-based precipitation nowcasting,

G. Ayzel, T. Scheffer, and M. Heistermann, "RainNet v1.0: a convolutional neural network for radar-based precipitation nowcasting," Geoscientific Model Development, vol. 13, no. 6, pp. 2631-2644, 2020

2020

[21] [21]

Precipitation nowcasting using transformer-based generative models and transfer learning for improved disaster preparedness,

M. J. Piran, X. Wang, H. J. Kim, and H. H. Kwon, "Precipitation nowcasting using transformer-based generative models and transfer learning for improved disaster preparedness," International Journal of Applied Earth Observation and Geoinformation, vol. 132, 2024

2024

[22] [22]

Environmental monitoring by wireless communication networks,

H. Messer, A. Zinevich, and P. Alpert, "Environmental monitoring by wireless communication networks," Science, vol. 312, no. 5774, p. 713, 2006

2006

[23] [23]

Rainfall measurement using radio links from cellular communication networks,

H. Leijnse, R. Uijlenhoet, and J. N. M. Stricker, "Rainfall measurement using radio links from cellular communication networks," Water Resources Research, vol. 43, no. 3, 2007

2007

[24] [24]

Rain event detection in commercial microwave link attenuation data using convolutional neural networks,

J. Polz, C. Chwala, M. Graf, and H. Kunstmann, "Rain event detection in commercial microwave link attenuation data using convolutional neural networks," Atmospheric Measurement Techniques, vol. 13, no. 7, pp. 3835-3853, 2020

2020

[25] [25]

Commercial microwave link networks for rainfall observation: Assessment of the current status and future challenges,

C. Chwala and H. Kunstmann, "Commercial microwave link networks for rainfall observation: Assessment of the current status and future challenges," WIREs Water, vol. 6, no. 2, 2019

2019

[26] [26]

A Review on Rainfall Measurement Based on Commercial Microwave Links in Wireless Cellular Networks,

B. Lian, Z. Wei, X. Sun, Z. Li, and J. Zhao, "A Review on Rainfall Measurement Based on Commercial Microwave Links in Wireless Cellular Networks," Sensors (Basel), vol. 22, no. 12, Jun 10 2022

2022

[27] [27]

Rain Attenuation Statistics for a 120-GHz-Band Wireless Link,

A. Hirata, H. Takahashi, T. Kosugi, K. Murata, K. Naoya, and Y. Kado, "Rain Attenuation Statistics for a 120-GHz-Band Wireless Link," Ieee Mtt S Int Micr, pp. 933, 2009

2009

[28] [28]

Specific attenuation model for rain for use in prediction methods,

ITU-R, "Specific attenuation model for rain for use in prediction methods," Geneva, Switzerland, Recommendation ITU-R P.838-3, 2005

2005

[29] [29]

Statistical theory of the Marshall-Palmer distribution of raindrops,

L. Yangang, "Statistical theory of the Marshall-Palmer distribution of raindrops," Atmospheric Environment. Part A. General Topics, vol. 27, no. 1, pp. 15-19, 1993

1993

[30] [30]

Feloni, K

E. Feloni, K. Kotsifakis, N. Dervos, G. Giavis, and E. Baltas, Analysis of Joss-Waldvogel disdrometer measurements in rainfall events (Fifth International Conference on Remote Sensing and Geoinformation of the Environment (RSCy2017)). SPIE, 2017

2017

[31] [31]

Drop-Size Distributions and Mie Computations for Rain,

C. Mätzler, "Drop-Size Distributions and Mie Computations for Rain," Institut für Angewandte Physik, Universität Bern, Bern, Switzerland, November 2002 2002

2002

[32] [32]

Microwave link dual-wavelength measurements of path-average attenuation for the estimation of drop size distributions and rainfall,

R. F. Rincon and R. H. Lang, "Microwave link dual-wavelength measurements of path-average attenuation for the estimation of drop size distributions and rainfall," IEEE Transactions on Geoscience and Remote Sensing, vol. 40, no. 4, pp. 760-770, 2002

2002

[33] [33]

Estimating raindrop size distributions using microwave link measurements: potential and limitations,

T. C. van Leth, H. Leijnse, A. Overeem, and R. Uijlenhoet, "Estimating raindrop size distributions using microwave link measurements: potential and limitations," Atmospheric Measurement Techniques, vol. 13, no. 4, pp. 1797-1815, 2020

2020

[34] [34]

Measurements and models for 38-GHz point-to-multipoint radiowave propagation,

X. Hao, T. S. Rappaport, R. J. Boyle, and J. H. Schaffner, "Measurements and models for 38-GHz point-to-multipoint radiowave propagation," IEEE Journal on Selected Areas in Communications, vol. 18, no. 3, pp. 310-321, 2000

2000

[35] [35]

Bootstrap Methods: Another Look at the Jackknife,

B. Efron, "Bootstrap Methods: Another Look at the Jackknife," The Annals of Statistics, vol. 7, no. 1, 1979

1979

[36] [36]

Time- and Frequency-Varying K-Factor of Non-Stationary Vehicular Channels for Safety-Relevant Scenarios,

L. Bernado, T. Zemen, F. Tufvesson, A. F. Molisch, and C. F. Mecklenbrauker, "Time- and Frequency-Varying K-Factor of Non-Stationary Vehicular Channels for Safety-Relevant Scenarios," IEEE Transactions on Intelligent Transportation Systems, vol. 16, no. 2, pp. 1007-1017, 2015

2015

[37] [37]

On the estimation of the K parameter for the Rice fading distribution,

A. Abdi, C. Tepedelenlioglu, M. Kaveh, and G. Giannakis, "On the estimation of the K parameter for the Rice fading distribution," IEEE Communications Letters, vol. 5, no. 3, pp. 92-94, 2001

2001

[38] [38]

Rain: relaxations in the sky,

O. Peters and K. Christensen, "Rain: relaxations in the sky," Phys Rev E Stat Nonlin Soft Matter Phys, vol. 66, no. 3 Pt 2A, p. 036120, Sep 2002

2002

[39] [39]

Understanding and improving layer normalization

J. Xu, et al. "Understanding and improving layer normalization." presented at the Advances in neural information processing systems 32, 2019

2019

[40] [40]

Attention is All you Need,

A. Vaswani et al., "Attention is All you Need," presented at the Advances in Neural Information Processing Systems 30, 2017

2017

[41] [41]

Language Modeling with Gated Convolutional Networks,

Y. N. Dauphin, A. Fan, M. Auli, and D. Grangier, "Language Modeling with Gated Convolutional Networks," presented at the Proceedings of the 34th International Conference on Machine Learning, 2017

2017

[42] [42]

Dropout: A Simple Way to Prevent Neural Networks from Overfitting,

N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, "Dropout: A Simple Way to Prevent Neural Networks from Overfitting," (in English), J Mach Learn Res, vol. 15, pp. 1929-1958, Jun 2014

1929

[43] [43]

A time series forest for classification and feature extraction,

H. Deng, G. Runger, E. Tuv, and M. Vladimir, "A time series forest for classification and feature extraction," Information Sciences, vol. 239, pp. 142-153, 2013

2013

[44] [44]

Domain randomization for transferring deep neural networks from simulation to the real world,

J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel, "Domain randomization for transferring deep neural networks from simulation to the real world," presented at the 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017

2017

[45] [45]

Physics-informed machine learning,

G. E. Karniadakis, I. G. Kevrekidis, L. Lu, P. Perdikaris, S. Wang, and L. Yang, "Physics-informed machine learning," Nature Reviews Physics, vol. 3, no. 6, pp. 422-440, 2021

2021

[46] [46]

Are Transformers Effective for Time Series Forecasting?,

A. Zeng, M. Chen, L. Zhang, and Q. Xu, "Are Transformers Effective for Time Series Forecasting?," Proceedings of the AAAI Conference on Artificial Intelligence, vol. 37, no. 9, pp. 11121-11128, 2023

2023

[47] [47]

Radar Reflectivity and Meteorological Factors Merging‐Based Precipitation Estimation Neural Network,

Y. Zhang, S. Chen, W. Tian, and S. Chen, "Radar Reflectivity and Meteorological Factors Merging‐Based Precipitation Estimation Neural Network," Earth and Space Science, vol. 8, no. 10, 2021

2021

[48] [48]

Deep learning for an improved prediction of rainfall retrievals from commercial microwave links

J. Pudashine, et al., "Deep learning for an improved prediction of rainfall retrievals from commercial microwave links." Water Resources Research 56, no. 7, pp. e2019WR026255, 2020

2020

[49] [49]

Transformers in Time Series: A Survey,

Q. Wen et al., "Transformers in Time Series: A Survey," presented at the Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, 2023

2023

[50] [50]

A Time Series Is Worth 64 Words: Long-term Forecasting with Transformers,

Y. Nie, N. H. Nguyen, P. Sinthong, and J. Kalagnanam, "A Time Series Is Worth 64 Words: Long-term Forecasting with Transformers," presented at the International Conference on Learning Representations, 2023

2023

[51] [51]

Root mean square error (RMSE) or mean absolute error (MAE)? – Arguments against avoiding RMSE in the literature,

T. Chai and R. R. Draxler, "Root mean square error (RMSE) or mean absolute error (MAE)? – Arguments against avoiding RMSE in the literature," Geoscientific Model Development, vol. 7, no. 3, pp. 1247-1250, 2014

2014

[52] [52]

Conformer: Convolution-augmented Transformer for Speech Recognition,

A. Gulati et al., "Conformer: Convolution-augmented Transformer for Speech Recognition," presented at the Interspeech 2020, 2020

2020