arxiv: 2604.17414 · v1 · submitted 2026-04-19 · 📡 eess.SP

Recognition: unknown

Physics-Aware Query-Conditioned Graph Attention Networks for Radio Map Estimation

Ang Li , Chengyu Liu , Yue Wang

Authors on Pith no claims yet

Pith reviewed 2026-05-10 06:05 UTC · model grok-4.3

classification 📡 eess.SP

keywords radio map estimationgraph attention networkssparse measurementswireless signal strengthquery-conditioned estimationphysics-aware learningDeepMIMO

0 comments

The pith

A physics-aware hierarchical graph attention network estimates transmitter-specific signal strength at queried locations from sparse measurements using only geometric descriptors.

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

The paper proposes a method for estimating received signal strength at arbitrary points relative to a transmitter by building a local graph that connects each query to nearby reference observations and aggregates information with transmitter-referenced distances and angles. A second global graph then shares context across nearby queries to enforce neighborhood consistency. This matters because wireless planning often requires estimates only at specific locations or quick updates after local changes rather than full-grid reconstruction, and the approach avoids any need for building layouts or other environment data. Three operating modes let the network either predict directly or improve upon an existing rough prior. Simulations on the DeepMIMO dataset show the direct mode beats other learning baselines in error metrics, while the refinement modes cut prior error further.

Core claim

The proposed physics-aware query-conditioned hierarchical graph attention network constructs a bounded local graph over sampled reference observations for each queried target-transmitter pair, aggregates reference-to-query evidence through transmitter-referenced geometric descriptors, and employs a global graph to exchange representation-level context among nearby target locations. This architecture supports three operating regimes: direct RSS estimation, prior-conditioned residual correction, and post-hoc gated attenuation of the learned correction. The framework relies solely on measurement-side quantities without environment-side inputs. On the DeepMIMO scenario, the direct regime of the

What carries the argument

Hierarchical graph attention network (HGAT) that builds a local graph around each query-transmitter pair to aggregate evidence via transmitter-referenced geometric descriptors and then uses a global graph for neighborhood consistency.

Where Pith is reading between the lines

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

The query-based design could support efficient localized map updates by re-estimating only around changed areas rather than rebuilding everything.
The same local-plus-global graph structure might transfer to other wave-propagation tasks such as acoustic or optical field estimation if the geometric descriptors are adapted.
Because the model needs no environment inputs, it could be deployed on edge devices that only collect measurements on demand.

Load-bearing premise

The DeepMIMO ray-tracing simulations sufficiently represent the propagation physics of real deployments, and the chosen transmitter-referenced geometric descriptors capture the necessary effects without explicit environment modeling or post-hoc tuning.

What would settle it

Run the trained HGAT on a set of sparse real-world RSS measurements collected in an actual urban or indoor site with known transmitter locations, then compare its RMSE and MAE against the same baselines and any available prior estimates.

Figures

Figures reproduced from arXiv: 2604.17414 by Ang Li, Chengyu Liu, Yue Wang.

**Figure 2.** Figure 2: Illustration of the HGAT encoder architecture operating in the direct regime. A sampled reference scaffold provides a reusable evidence set for each [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Figure 3: Block diagram illustration of workflows of the three operating regimes. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

**Figure 4.** Figure 4: Representative heatmap comparison on Denver Site [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗

**Figure 5.** Figure 5: Representative heatmap comparison on Denver Site [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗

**Figure 6.** Figure 6: Residual-regime qualitative comparison on Denver Site 3 (seen) with [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗

**Figure 7.** Figure 7: Post-hoc gate diagnostic on Denver Site 3 (seen). The learned [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗

read the original abstract

Radio map estimation from sparse measurements is fundamental to wireless network planning, optimization, and localized map updating. Most recent learning-based approaches formulate the problem as dense map completion over a predefined grid, whereas many practical deployments require estimating transmitter-specific received signal strength only at queried locations or refining an existing map after local changes. This paper proposes a physics-aware query-conditioned hierarchical graph attention network for transmitter-resolved point-wise radio map estimation. For each queried target--transmitter pair, the proposed encoder constructs a bounded local graph over sampled reference observations and aggregates reference-to-query evidence through transmitter-referenced geometric descriptors. A global graph then exchanges representation-level context among nearby target locations to improve neighborhood consistency without revisiting a large number of reference measurements. On top of this shared architecture, we instantiate three operating regimes: direct RSS estimation, prior-conditioned residual correction, and post-hoc gated attenuation of the learned correction. The framework uses only measurement-side quantities and does not rely on environment-side inputs. Simulations on the DeepMIMO scenario show that, in the direct regime, the proposed HGAT achieves the lowest RMSE and MAE among the evaluated learning-based baselines on all reported sites. When conventional prior estimate is available, the residual and gated regimes further reduce the prior error.

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

This paper gives a workable hierarchical GAT for point-wise radio map queries that beats baselines in DeepMIMO simulations, but the gains rest only on those simulations without ablations or real data.

read the letter

The one thing to know is that this paper introduces a hierarchical graph attention network for estimating radio signal strength at specific query points rather than filling entire grids, and it shows better accuracy than the baselines in their DeepMIMO tests. The architecture is new in how it conditions the graph on the query and transmitter pair. It creates a local graph with reference measurements, using descriptors based on distances and angles to the transmitter to bring in some physics. Then a global graph connects the query points to enforce consistency in neighborhoods. On top, they have direct estimation, or ways to refine an existing map with residuals or gating. This setup does a good job in the simulations. In the direct regime, it has the lowest errors among the learning-based methods across the sites. When priors are used, the other regimes improve on them too. The fact that it only needs measurement data and no full environment model is a plus for practicality. Where it is softer is the evaluation. There are no ablations to isolate the contribution of the hierarchical graphs or the descriptors. The results have no error bars, so we do not know the variability. Everything is on ray-tracing simulations, which may not fully match real deployments with their complexities. The choice of regime seems post-hoc, which could affect how general the gains are. This paper is for engineers and researchers focused on wireless network planning and optimization who deal with sparse sensor data. It could also interest those applying graph networks to inverse problems in physics. It deserves a serious referee because the core idea is well-motivated and the simulation results support the claims within their scope. The method is described clearly enough to reproduce. I would send it for peer review, suggesting they add ablations, statistical analysis, and some real-world data if possible to make the case stronger.

Referee Report

1 major / 2 minor

Summary. The paper proposes a physics-aware query-conditioned hierarchical graph attention network (HGAT) for transmitter-resolved point-wise radio map estimation from sparse measurements. It builds bounded local graphs using transmitter-referenced geometric descriptors to aggregate reference-to-query evidence, followed by a global graph for neighborhood consistency, and supports three regimes (direct RSS estimation, prior-conditioned residual correction, and post-hoc gated attenuation). The framework operates solely on measurement-side quantities. DeepMIMO ray-tracing simulations show that the direct-regime HGAT achieves the lowest RMSE and MAE among evaluated learning-based baselines across reported sites, with residual and gated regimes further reducing prior error when available.

Significance. If the performance ordering holds under more rigorous validation, the work provides a practical, query-specific alternative to dense grid-completion methods for radio map tasks, with built-in support for incorporating priors via residual/gated modes and explicit use of geometric descriptors to encode propagation physics without environment-side inputs. The hierarchical local-to-global graph design and measurement-only constraint are notable strengths for deployment flexibility.

major comments (1)

[Experimental results on DeepMIMO scenario] Experimental results (DeepMIMO evaluation): the claim of lowest RMSE/MAE in the direct regime across sites lacks error bars, standard deviations from repeated trials, ablation studies on graph construction or attention components, or statistical significance tests against baselines; without these, the reported gains cannot be assessed for robustness or generalizability beyond the specific simulation protocol.

minor comments (2)

[Abstract and §3 (method)] The abstract and method description would benefit from a concise diagram or pseudocode clarifying the data flow between the local graph encoder, global graph, and the three operating regimes.
[§3.1] Notation for geometric descriptors and graph construction could be made more explicit (e.g., explicit definition of bounded local graph radius) to aid reproducibility.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback and the recommendation of minor revision. We address the major comment on the experimental results point by point below.

read point-by-point responses

Referee: Experimental results (DeepMIMO evaluation): the claim of lowest RMSE/MAE in the direct regime across sites lacks error bars, standard deviations from repeated trials, ablation studies on graph construction or attention components, or statistical significance tests against baselines; without these, the reported gains cannot be assessed for robustness or generalizability beyond the specific simulation protocol.

Authors: We agree that the current presentation of results would benefit from additional statistical rigor to better support claims of robustness and generalizability. In the revised manuscript we will rerun all DeepMIMO experiments across multiple independent random seeds (reporting mean and standard deviation for RMSE and MAE), include ablation studies that isolate the contributions of the bounded local-graph construction, transmitter-referenced geometric descriptors, and hierarchical attention layers, and add statistical significance tests (e.g., paired t-tests or Wilcoxon signed-rank tests) against the baselines. Updated tables and figures will be provided. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper introduces a hierarchical graph attention network architecture for point-wise radio map estimation from sparse measurements, using transmitter-referenced geometric descriptors and query-conditioned graphs. All performance claims rest on empirical training and evaluation against external DeepMIMO ray-tracing simulations and a set of independent learning-based baselines. No equations, predictions, or uniqueness arguments reduce by construction to fitted parameters, self-definitions, or self-citation chains; the reported RMSE/MAE ordering follows directly from the described experimental protocol on held-out simulation data.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The claim rests on the assumption that geometric descriptors and attention suffice to encode propagation physics, with performance demonstrated only via simulation; no new physical entities or axioms beyond standard ML training are introduced.

free parameters (1)

model parameters and hyperparameters
Neural network weights are fitted to training data from DeepMIMO; this is standard but means results depend on training procedure and data splits not detailed in the abstract.

axioms (1)

domain assumption DeepMIMO ray-tracing simulations provide representative data for real radio propagation
All reported performance gains rely on this dataset being sufficient to validate superiority over baselines.

pith-pipeline@v0.9.0 · 5516 in / 1239 out tokens · 62373 ms · 2026-05-10T06:05:05.292461+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

45 extracted references · 2 canonical work pages

[1]

Spectrum surveying: Active radio map estimation with autonomous UA Vs,

R. Shrestha, D. Romero, and S. P. Chepuri, “Spectrum surveying: Active radio map estimation with autonomous UA Vs,”IEEE Trans. Wireless Commun., vol. 22, no. 1, pp. 627–641, Jan. 2023

2023
[2]

Radio map estimation: A data-driven approach to spectrum cartography,

D. Romero and S.-J. Kim, “Radio map estimation: A data-driven approach to spectrum cartography,”IEEE Signal Process. Mag., vol. 39, no. 6, pp. 53–72, Oct. 2022

2022
[3]

Aerial base station placement leveraging radio tomographic maps,

D. Romero, P. Q. Viet, and G. Leus, “Aerial base station placement leveraging radio tomographic maps,” inIEEE Int. Conf. Acoust., Speech Signal Process. (ICASSP), 2022, pp. 5358–5362

2022
[4]

Engineering radio maps for wireless resource management,

S. Bi, J. Lyu, Z. Ding, and R. Zhang, “Engineering radio maps for wireless resource management,”IEEE Wireless Commun., vol. 26, no. 2, pp. 133–141, 2019

2019
[5]

Real-time outdoor localization using radio maps: A deep learning approach,

c. Yapar, R. Levie, G. Kutyniok, and G. Caire, “Real-time outdoor localization using radio maps: A deep learning approach,”IEEE Trans. Wireless Commun., vol. 22, no. 12, pp. 9703–9717, Dec. 2023. 12

2023
[6]

Rem-based channel awareness in time-varying environments for CSI feedback reduction,

F. Burmeister, R. Walstab, A. Sch ¨osser, M. Matth´e, P. Schulz, and G. Fet- tweis, “Rem-based channel awareness in time-varying environments for CSI feedback reduction,” inIEEE Int. Conf. Commun., 2025, pp. 1912– 1917

2025
[7]

Theoretical analysis of the radio map estimation problem,

D. Romero, T. N. Ha, R. Shrestha, and M. Franceschetti, “Theoretical analysis of the radio map estimation problem,”IEEE Trans. Wireless Commun., vol. 23, no. 10, pp. 13 722–13 737, May 2024

2024
[8]

An adaptive inverse-distance weighting spatial interpolation technique,

G. Y . Lu and D. W. Wong, “An adaptive inverse-distance weighting spatial interpolation technique,”Comput. Geosci., vol. 34, no. 9, pp. 1044–1055, 2008. [Online]. Available: https://www.sciencedirect.com/ science/article/pii/S0098300408000721

2008
[9]

Fast construction of the radio map based on the improved low-rank matrix completion and recovery method for an indoor positioning system,

Z. Wang, L. Zhang, Q. Kong, and K. Wang, “Fast construction of the radio map based on the improved low-rank matrix completion and recovery method for an indoor positioning system,”J. Sensors, vol. 2021, no. 1, p. 2017208, 2021

2021
[10]

Propagation map reconstruction via interpolation assisted matrix completion,

H. Sun and J. Chen, “Propagation map reconstruction via interpolation assisted matrix completion,”IEEE Trans. Signal Process., vol. 70, pp. 6154–6169, 2022

2022
[11]

Integrated interpolation and matrix completion for radio map estimation: A convex optimization approach,

H. Dong, W. Pu, R. Zhou, X. Fu, and F. Yin, “Integrated interpolation and matrix completion for radio map estimation: A convex optimization approach,” inIEEE Int. Conf. Acoust., Speech Signal Process. (ICASSP), 2025, pp. 1–5

2025
[12]

Kriging: a method of interpolation for geographical information systems,

M. A. OLIVER and R. WEBSTER, “Kriging: a method of interpolation for geographical information systems,”Int. J. Geograph. Inf. Syst., vol. 4, no. 3, pp. 313–332, 1990

1990
[13]

Ordinary kriging,

H. Wackernagel, “Ordinary kriging,” inMultivariate geostatistics: an introduction with applications. Springer, 2003, pp. 79–88

2003
[14]

Learn- ing power spectrum maps from quantized power measurements,

D. Romero, S.-J. Kim, G. B. Giannakis, and R. L ´opez-Valcarce, “Learn- ing power spectrum maps from quantized power measurements,”IEEE Trans. Signal Process., vol. 65, no. 10, pp. 2547–2560, May 2017

2017
[15]

Non-parametric spectrum cartog- raphy using adaptive radial basis functions,

M. Hamid and B. Beferull-Lozano, “Non-parametric spectrum cartog- raphy using adaptive radial basis functions,” inIEEE Int. Conf. Acoust., Speech Signal Process. (ICASSP), 2017, pp. 3599–3603

2017
[16]

RecNet: A convolutional network for efficient radiomap reconstruction,

Q. Niu, Y . Nie, S. He, N. Liu, and X. Luo, “RecNet: A convolutional network for efficient radiomap reconstruction,” in2018 IEEE Int. Conf. Commun. (ICC), 2018, pp. 1–7

2018
[17]

Transfer learning for tilt-dependent radio map prediction,

C. Parera, Q. Liao, I. Malanchini, C. Tatino, A. E. C. Redondi, and M. Cesana, “Transfer learning for tilt-dependent radio map prediction,” IEEE Trans. Cogn. Commun. Netw., vol. 6, no. 2, pp. 829–843, Jun. 2020

2020
[18]

FadeNet: Deep learning-based mm- wave large-scale channel fading prediction and its applications,

V . V . Ratnam, H. Chen, S. Pawar, B. Zhang, C. J. Zhang, Y .-J. Kim, S. Lee, M. Cho, and S.-R. Yoon, “FadeNet: Deep learning-based mm- wave large-scale channel fading prediction and its applications,”IEEE Access, vol. 9, pp. 3278–3290, Dec. 2021

2021
[19]

RadioUNet: Fast radio map estimation with convolutional neural networks,

R. Levie, C ¸ . Yapar, G. Kutyniok, and G. Caire, “RadioUNet: Fast radio map estimation with convolutional neural networks,”IEEE Trans. Wireless Commun., vol. 20, no. 6, pp. 4001–4015, Feb. 2021

2021
[20]

Deep completion autoencoders for radio map estimation,

Y . Teganya and D. Romero, “Deep completion autoencoders for radio map estimation,”IEEE Trans. Wireless Commun., vol. 21, no. 3, pp. 1710–1724, Aug. 2022

2022
[21]

Deep-learning-based radio map reconstruction for V2X communications,

S. Roger, M. Brambilla, B. C. Tedeschini, C. Botella-Mascarell, M. Co- bos, and M. Nicoli, “Deep-learning-based radio map reconstruction for V2X communications,”IEEE Trans. Veh. Technol., vol. 73, no. 3, pp. 3863–3871, Mar. 2024

2024
[22]

A graph neural network based radio map construction method for urban environment,

G. Chen, Y . Liu, T. Zhang, J. Zhang, X. Guo, and J. Yang, “A graph neural network based radio map construction method for urban environment,”IEEE Commun. Lett., vol. 27, no. 5, pp. 1327–1331, May 2023

2023
[23]

RME-GAN: A learning framework for radio map estimation based on conditional generative adversarial network,

S. Zhang, A. Wijesinghe, and Z. Ding, “RME-GAN: A learning framework for radio map estimation based on conditional generative adversarial network,”IEEE Int. Things J., vol. 10, no. 20, pp. 18 016– 18 027, Oct. 2023

2023
[24]

Two-stage radio map construction with real environments and sparse measurements,

Y . Wang, S. Sun, N. Liu, L. Xu, and L. Wang, “Two-stage radio map construction with real environments and sparse measurements,”IEEE Wireless Commun. Lett., vol. 14, no. 4, pp. 969–973, Jan. 2025

2025
[25]

Dual-GRE: Dual-phase enhancement in radiomap estimation based on graph attention,

Y . Ma, Y . Zhou, S. Zhang, and Z. Ding, “Dual-GRE: Dual-phase enhancement in radiomap estimation based on graph attention,”IEEE Wireless Commun. Lett., vol. 14, no. 8, pp. 2646–2650, Aug. 2025

2025
[26]

RadioGAT: A joint model-based and data-driven framework for multi-band radiomap reconstruction via graph attention networks,

X. Li, S. Zhang, H. Li, X. Li, L. Xu, H. Xu, H. Mei, G. Zhu, N. Qi, and M. Xiao, “RadioGAT: A joint model-based and data-driven framework for multi-band radiomap reconstruction via graph attention networks,” IEEE Trans. Wireless Commun., vol. 23, no. 11, pp. 17 777–17 792, Sep. 2024

2024
[27]

On the performance of neural network residual kriging in radio environment mapping,

K. Sato, K. Inage, and T. Fujii, “On the performance of neural network residual kriging in radio environment mapping,”IEEE Access, vol. 7, pp. 94 557–94 568, 2019

2019
[28]

Kriging convolutional networks,

G. Appleby, L. Liu, and L.-P. Liu, “Kriging convolutional networks,” Proceed. AAAI Conf. Artif. Intell., vol. 34, no. 04, pp. 3187–3194, Apr
[29]

Available: https://ojs.aaai.org/index.php/AAAI/article/ view/5716

[Online]. Available: https://ojs.aaai.org/index.php/AAAI/article/ view/5716
[30]

Spatial transformers for radio map estima- tion,

P. Q. Viet and D. Romero, “Spatial transformers for radio map estima- tion,” inIEEE Int. Conf. Commun., 2025, pp. 6155–6160

2025
[31]

Deep spectrum cartography: Com- pleting radio map tensors using learned neural models,

S. Shrestha, X. Fu, and M. Hong, “Deep spectrum cartography: Com- pleting radio map tensors using learned neural models,”IEEE Trans. Signal Process., vol. 70, pp. 1170–1184, Jan. 2022

2022
[32]

Inductive graph neural networks for spatiotemporal kriging,

Y . Wu, D. Zhuang, A. Labbe, and L. Sun, “Inductive graph neural networks for spatiotemporal kriging,”Proc. AAAI Conf. Artif. Intell., vol. 35, no. 5, pp. 4478–4485, May 2021. [Online]. Available: https://ojs.aaai.org/index.php/AAAI/article/view/16575

2021
[33]

Radio map interpolation using graph signal process- ing,

A. E. C. Redondi, “Radio map interpolation using graph signal process- ing,”IEEE Commun. Lett., vol. 22, no. 1, pp. 153–156, Oct. 2018

2018
[34]

A tutorial on environment-aware communications via channel knowledge map for 6G,

Y . Zeng, J. Chen, J. Xu, D. Wu, X. Xu, S. Jin, X. Gao, D. Gesbert, S. Cui, and R. Zhang, “A tutorial on environment-aware communications via channel knowledge map for 6G,”IEEE Commun. Surv. Tut., vol. 26, no. 3, pp. 1478–1519, Feb. 2024

2024
[35]

Radio measurement collection for Minimization of Drive Tests (MDT); Overall description; Stage 2,

3GPP, “Radio measurement collection for Minimization of Drive Tests (MDT); Overall description; Stage 2,” 3GPP TS 37.320, available from the 3GPP specification portal
[36]

System architecture for the 5G System (5GS),

——, “System architecture for the 5G System (5GS),” 3GPP TS 23.501, available from the 3GPP specification portal
[37]

5GNN: extrapolating 5G measurements through GNNs,

W. Ye, X. Hu, T. Liu, R. Sun, Y . Li, and Z.-L. Zhang, “5GNN: extrapolating 5G measurements through GNNs,” inProc. 1st Int. Workshop Graph Neural Netw., ser. GNNet ’22. New York, NY , USA: Association for Computing Machinery, 2022, p. 36–41. [Online]. Available: https://doi.org/10.1145/3565473.3569186

work page doi:10.1145/3565473.3569186 2022
[38]

Analysis of interpolation errors in urban digital surface models created from LIDAR data,

G. Gonc ¸alves, “Analysis of interpolation errors in urban digital surface models created from LIDAR data,” in7th Int. Symp. spatial accuracy assessment in natural resources and environmental sciences, 2006, pp. 160–168

2006
[39]

Vari- ational approaches on discontinuity localization and field estimation in sea surface temperature and soil moisture,

W. Sun, M. C ¸ etin, W. C. Thacker, T. M. Chin, and A. S. Willsky, “Vari- ational approaches on discontinuity localization and field estimation in sea surface temperature and soil moisture,”IEEE Trans. Geosci. Remote Sens., vol. 44, no. 2, pp. 336–350, 2006

2006
[40]

Inductive representation learning on large graphs,

W. Hamilton, Z. Ying, and J. Leskovec, “Inductive representation learning on large graphs,” inAdvances Neural Inf. Pross. Syst., I. Guyon, U. V . Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, Eds., vol. 30. Curran Associates, Inc., 2017. [Online]. Available: https://proceedings.neurips.cc/paper files/paper/2017/file/5dd9db5e0...

2017
[41]

GraphSAINT: Graph sampling based inductive learning method,

H. Zeng, H. Zhou, A. Srivastava, R. Kannan, and V . Prasanna, “GraphSAINT: Graph sampling based inductive learning method,” inInt. Conf. Learn. Representations, 2020. [Online]. Available: https://openreview.net/forum?id=BJe8pkHFwS

2020
[42]

FastGCN: Fast learning with graph convolutional networks via importance sampling,

J. Chen, T. Ma, and C. Xiao, “FastGCN: Fast learning with graph convolutional networks via importance sampling,” inInt. Conf. Learn. Representations, 2018. [Online]. Available: https: //openreview.net/forum?id=rytstxW AW

2018
[43]

DeepMIMO: A generic deep learning dataset for mil- limeter wave and massive MIMO applications,

A. Alkhateeb, “DeepMIMO: A generic deep learning dataset for mil- limeter wave and massive MIMO applications,” inProc. Inf. Theory Appl. Workshop (ITA), San Diego, CA, Feb 2019, pp. 1–8

2019
[44]

Exploring the limits of transfer learning with a unified text-to-text transformer,

C. Raffel, N. Shazeer, A. Roberts, K. Lee, S. Narang, M. Matena, Y . Zhou, W. Li, and P. J. Liu, “Exploring the limits of transfer learning with a unified text-to-text transformer,”J. Mach. Learn. Res., vol. 21, no. 140, pp. 1–67, 2020. [Online]. Available: http://jmlr.org/papers/v21/20-074.html

2020
[45]

URL https://doi.org/10.1145/3528223

T. M ¨uller, A. Evans, C. Schied, and A. Keller, “Instant neural graphics primitives with a multiresolution hash encoding,”ACM Trans. Graph., vol. 41, no. 4, Jul. 2022. [Online]. Available: https://doi.org/10.1145/3528223.3530127

work page doi:10.1145/3528223.3530127 2022