pith. sign in

arxiv: 2607.01453 · v1 · pith:RTR74LZ3new · submitted 2026-07-01 · 📡 eess.SP

Channel Knowledge Map Reconstruction From Sparse Measurements via Pilot-Anchored Layout-Conditioned Fourier Refinement

Zhonghao Jiu , Fan Meng , Yongming Huang , Hang Zhan , Zening Liu , Xiaohu You This is my paper

Pith reviewed 2026-07-03 18:28 UTC · model grok-4.3

classification 📡 eess.SP
keywords channel knowledge mapsparse reconstructionpartial convolutionFourier refinementlayout conditioningwireless channel mappingDeepMIMO
0
0 comments X

The pith

Anchor-CKM reconstructs channel knowledge maps from sparse measurements by first building a pilot-supported representation before layout-conditioned Fourier refinement.

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

The paper establishes that applying local aggregation or spectral operators directly to zero-filled pilot grids entangles the sampling mask with the channel field, allowing structural priors to act on mask-induced distortions. Anchor-CKM addresses this by first using support-aware partial convolutions to construct a pilot-supported representation, then performing layout-conditioned dual-path Fourier refinement and coordinate-based heteroscedastic prediction of the CKM mean and variance. Experiments on transmitter-disjoint DeepMIMO scenarios with missing ratios from 0.3 to 0.95 show received-power RMSE reductions of 0.79 to 1.33 dB relative to the strongest baseline in explicit-layout outdoor cases. Ablations identify pilot-support stabilization as the largest contributor and layout conditioning as beneficial for LOS/NLOS boundary fidelity. A sympathetic reader would care because long-term environmental changes require periodic CKM refreshes from sparse irregular measurements when only coarse layout descriptors are available.

Core claim

Anchor-CKM is a measurement-first, knowledge-aided framework that first constructs a pilot-supported representation via support-aware partial convolutions, then applies layout-conditioned dual-path Fourier refinement followed by coordinate-based heteroscedastic prediction, yielding 0.79 to 1.33 dB lower received-power RMSE than reproduced baselines across missing ratios 0.3-0.95 in explicit-layout outdoor DeepMIMO scenarios.

What carries the argument

Support-aware partial convolutions to build pilot-supported representation, followed by layout-conditioned dual-path Fourier refinement and heteroscedastic prediction

If this is right

  • Pilot-support stabilization is the largest single contributor to the observed error reduction according to the ablations.
  • Layout conditioning improves reconstruction fidelity specifically at line-of-sight to non-line-of-sight boundaries.
  • The performance gains hold across missing ratios from 0.3 to 0.95, including stringent 5-10% pilot coverage settings.
  • The framework is evaluated on transmitter-disjoint scenarios to ensure the reconstruction does not rely on transmitter-specific artifacts.

Where Pith is reading between the lines

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

  • The ordering of operations (support stabilization before spectral refinement) may reduce mask entanglement in other sparse field reconstruction problems outside wireless channels.
  • If layout descriptors become outdated faster than expected, the boundary-fidelity benefit would diminish first.
  • The heteroscedastic variance output could be used downstream to guide where additional measurements should be collected.

Load-bearing premise

The method assumes that coarse layout or topology descriptors are available and sufficiently accurate to condition the dual-path Fourier refinement without introducing new distortions at LOS/NLOS boundaries.

What would settle it

Re-running the DeepMIMO experiments with the pilot-support stabilization step removed would show whether the full reported RMSE reduction persists or shrinks substantially.

Figures

Figures reproduced from arXiv: 2607.01453 by Fan Meng, Hang Zhan, Xiaohu You, Yongming Huang, Zening Liu, Zhonghao Jiu.

Figure 1
Figure 1. Figure 1: CKM refresh with sparse probing. Blue cells denote pilot probes; [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Anchor-CKM architecture. Stage 1 caches layout-conditioned FiLM fields; Stage 2 applies PConv-DtS densification and FiLM-modulated C-FNO [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Fourier compressibility on O1-60: representative CKM and sorted-mode cumulative energy. Bands show tile percentiles; dashed line marks [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Boundary-band error analysis on a representative O1-60 tile. Contour: 8-m near-building band; blue dots: pilots. Rows show mean-centered power, [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Missing-ratio robustness. RMSE and P90 are computed on unobserved valid cells for each scenario and the unweighted average. [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
read the original abstract

Channel knowledge maps (CKMs) enable environment-aware wireless systems by providing location-specific channel knowledge, but long-term environmental variations, such as construction, traffic redistribution, and foliage changes, require periodic map refresh. In practice, channel measurements are often sparse and irregular, while environmental knowledge may be limited to coarse layout or topology descriptors. This paper studies CKM reconstruction from sparse measurements. We show that reconstruction pipelines that apply local aggregation or spectral operators directly to a zero-filled pilot grid can entangle the sampling mask with the channel field, allowing structural priors to act on mask-induced distortions before the measurements define a supported radio field. To address this issue, we propose Anchor-CKM, a measurement-first, knowledge-aided reconstruction framework. Anchor-CKM first uses support-aware partial convolutions to construct a pilot-supported representation, and then performs layout-conditioned dual-path Fourier refinement followed by coordinate-based heteroscedastic prediction of the CKM mean and per-location predictive variance. Experiments on transmitter-disjoint DeepMIMO scenarios cover missing ratios from 0.3 to 0.95, including stringent 5% to 10% pilot-coverage settings. In explicit-layout outdoor scenarios, Anchor-CKM reduces received-power root-mean-square error (RMSE) by 0.79 to 1.33 dB relative to the strongest reproduced baseline, while ablations identify pilot-support stabilization as the largest contributor and layout conditioning as beneficial for line-of-sight/non-line-of-sight (LOS/NLOS) boundary fidelity.

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

1 major / 1 minor

Summary. The paper proposes Anchor-CKM, a measurement-first CKM reconstruction pipeline that first builds a pilot-supported representation via support-aware partial convolutions, then applies layout-conditioned dual-path Fourier refinement, and finally performs coordinate-based heteroscedastic prediction of mean and variance. On transmitter-disjoint DeepMIMO outdoor scenarios with missing ratios 0.3–0.95, it reports received-power RMSE reductions of 0.79–1.33 dB versus the strongest reproduced baseline, with ablations attributing the largest gain to pilot-support stabilization and secondary benefit to layout conditioning for LOS/NLOS boundary fidelity.

Significance. If the reported gains hold under realistic conditions, the framework offers a practical advance for periodic CKM refresh from sparse measurements by separating mask-induced artifacts from the radio field before applying structural priors. The explicit ablation results and focus on boundary fidelity provide concrete, falsifiable insights that could inform deployment in environment-aware systems.

major comments (1)
  1. [Abstract] Abstract (experiments paragraph): the central claim of 0.79–1.33 dB RMSE reduction in explicit-layout scenarios is measured only under presumably perfect coarse layouts; no sensitivity test or ablation on perturbed layout descriptors at LOS/NLOS boundaries is described, leaving open whether layout conditioning can entangle mask artifacts with layout-induced distortions and thereby undermine the reported benefit of the dual-path refinement step.
minor comments (1)
  1. [Abstract] The abstract states that experiments cover 'stringent 5% to 10% pilot-coverage settings' yet the missing-ratio range is given as 0.3–0.95; clarify the exact mapping between these quantities.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive comment on the experimental assumptions. We address the point below.

read point-by-point responses

  1. Referee: [Abstract] Abstract (experiments paragraph): the central claim of 0.79–1.33 dB RMSE reduction in explicit-layout scenarios is measured only under presumably perfect coarse layouts; no sensitivity test or ablation on perturbed layout descriptors at LOS/NLOS boundaries is described, leaving open whether layout conditioning can entangle mask artifacts with layout-induced distortions and thereby undermine the reported benefit of the dual-path refinement step.

    Authors: We agree that the reported gains assume accurate coarse layout descriptors as supplied by the DeepMIMO scenarios and that no sensitivity analysis on perturbed LOS/NLOS boundaries was performed. This is a genuine limitation of the presented experiments; the current results do not demonstrate robustness of the layout-conditioned refinement to layout descriptor errors. In the revision we will (i) qualify the abstract claim to read “under accurate coarse layout descriptors” and (ii) add a short discussion paragraph noting the untested sensitivity and its potential interaction with mask-induced artifacts. A quantitative ablation on perturbed layouts is noted as valuable future work but is not included in the present revision. revision: partial

Circularity Check

0 steps flagged

No significant circularity; derivation chain is self-contained with independent experimental validation

full rationale

The paper introduces Anchor-CKM as a new pipeline (support-aware partial convolutions followed by layout-conditioned dual-path Fourier refinement and heteroscedastic prediction) without any equations or claims that reduce reported RMSE gains to quantities defined by fitted parameters from the same data. Ablations are presented as identifying contributors (pilot-support stabilization as largest), but these are empirical breakdowns rather than self-definitional reductions. No self-citation chains, uniqueness theorems, or ansatzes smuggled via prior work are invoked in the provided text to force the central claims. The framework is externally falsifiable via DeepMIMO experiments across missing ratios, satisfying the criteria for a non-circular finding.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no explicit free parameters, axioms, or invented entities; the new components (partial convolutions, dual-path Fourier refinement) are algorithmic rather than postulated physical entities.

pith-pipeline@v0.9.1-grok · 5817 in / 1298 out tokens · 24847 ms · 2026-07-03T18:28:15.897968+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

43 extracted references · 43 canonical work pages · 2 internal anchors

  1. [1]

    When AI meets sustainable 6G,

    X. You, Y . Huang, C. Zhang, J. Wang, H. Yin, and H. Wu, “When AI meets sustainable 6G,”Sci. China Inf. Sci., vol. 68, no. 1, p. 110301, 2025

    work page 2025

  2. [2]

    PEMNet: Towards autonomous and enhanced environment-aware mobile networks,

    L. Li, Y . Xu, Y . Xue, F. Yin, C. Shen, R. Zhang, and T.-H. Chang, “PEMNet: Towards autonomous and enhanced environment-aware mobile networks,”arXiv preprint arXiv:2601.11025, 2026

    work page arXiv 2026

  3. [3]

    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. Surveys Tuts., vol. 26, no. 3, pp. 1478–1519, 2024

    work page 2024

  4. [4]

    Deep learning-based beam tracking for millimeter-wave communications under mobility,

    S. H. Lim, S. Kim, B. Shim, and J. W. Choi, “Deep learning-based beam tracking for millimeter-wave communications under mobility,”IEEE Trans. Commun., vol. 69, no. 11, pp. 7458–7469, Nov. 2021

    work page 2021

  5. [5]

    Channel knowledge map-assisted dual-domain tracking and predictive beamforming for high-mobility wireless networks,

    R. Du, Z. Wei, Z. Yang, L. Yang, Y . Zeng, D. W. K. Ng, and J. Yuan, “Channel knowledge map-assisted dual-domain tracking and predictive beamforming for high-mobility wireless networks,”arXiv preprint arXiv:2506.22796, 2025

    work page arXiv 2025

  6. [6]

    A recent survey on radio map estimation methods for wireless networks,

    B. Feng, M. Zheng, W. Liang, and L. Zhang, “A recent survey on radio map estimation methods for wireless networks,”Electronics, vol. 14, no. 8, p. 1564, 2025

    work page 2025

  7. [7]

    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,” 3rd Generation Partnership Project, Technical Specification TS 37.320, Jun. 2025, version 18.4.0, Release 18. [Online]. Available: https://www.3gpp.org/ftp/Specs/archive/ 37 series/37.320/37320-i40.zip

    work page 2025

  8. [8]

    Minimization of drive tests solution in 3GPP,

    W. A. Hapsari, A. Umesh, M. Iwamura, M. Tomala, B. Gyula, and B. S´ebire, “Minimization of drive tests solution in 3GPP,”IEEE Commun. Mag., vol. 50, no. 6, pp. 28–36, Jun. 2012

    work page 2012

  9. [9]

    The origins of kriging,

    N. Cressie, “The origins of kriging,”Math. Geol., vol. 22, no. 3, pp. 239–252, 1990

    work page 1990

  10. [10]

    A two-dimensional interpolation function for irregularly- spaced data,

    D. Shepard, “A two-dimensional interpolation function for irregularly- spaced data,” inProc. 1968 23rd ACM Nat. Conf.ACM Press, 1968, pp. 517–524

    work page 1968

  11. [11]

    Exact matrix completion via convex optimization,

    E. J. Cand `es and B. Recht, “Exact matrix completion via convex optimization,”Found. Comput. Math., vol. 9, no. 6, pp. 717–772, 2009

    work page 2009

  12. [12]

    Robust uncertainty principles: Ex- act signal reconstruction from highly incomplete frequency information,

    E. J. Cand `es, J. Romberg, and T. Tao, “Robust uncertainty principles: Ex- act signal reconstruction from highly incomplete frequency information,” IEEE Trans. Inf. Theory, vol. 52, no. 2, pp. 489–509, Feb. 2006

    work page 2006

  13. [13]

    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, Jun. 2021

    work page 2021

  14. [14]

    PMNet: Large-scale channel prediction system for ICASSP 2023 first pathloss radio map prediction challenge,

    J.-H. Lee, J. Lee, S.-H. Lee, and A. F. Molisch, “PMNet: Large-scale channel prediction system for ICASSP 2023 first pathloss radio map prediction challenge,” inProc. IEEE Int. Conf. Acoust., Speech, Signal Process. (ICASSP), 2023, pp. 1–2

    work page 2023

  15. [15]

    Graph neural networks for wireless communications: From theory to practice,

    Y . Shen, J. Zhang, S. Song, and K. B. Letaief, “Graph neural networks for wireless communications: From theory to practice,”IEEE Trans. Wireless Commun., vol. 22, no. 5, pp. 3554–3569, May 2023

    work page 2023

  16. [16]

    Principal neighbourhood aggregation for graph nets,

    G. Corso, L. Cavalleri, D. Beaini, P. Li `o, and P. Veli ˇckovi´c, “Principal neighbourhood aggregation for graph nets,” inProc. Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 33, 2020, pp. 13 260–13 271

    work page 2020

  17. [17]

    GraphMAE: Self-supervised masked graph autoencoders,

    Z. Hou, X. Liu, Y . Cen, Y . Dong, H. Yang, C. Wang, and J. Tang, “GraphMAE: Self-supervised masked graph autoencoders,” inProc. 28th ACM SIGKDD Conf. Knowl. Discov. Data Min., 2022, pp. 594–604

    work page 2022

  18. [18]

    RieIF: Knowledge-Driven Riemannian Information Flow for Robust Spatio-Temporal Graph Signal Prediction in 6G Wireless Networks

    Z. Jiu, Y . Huang, F. Meng, H. Zhan, Z. Liu, and X. You, “RieIF: Knowledge-driven riemannian information flow for robust spatio-temporal graph signal prediction in 6G wireless networks,”arXiv preprint arXiv:2604.06971, Apr. 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  19. [19]

    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, Nov. 2024

    work page 2024

  20. [20]

    RadioDiff: An effective generative diffusion model for sampling-free dynamic radio map construction,

    X. Wang, K. Tao, N. Cheng, Z. Yin, Z. Li, Y . Zhang, and X. Shen, “RadioDiff: An effective generative diffusion model for sampling-free dynamic radio map construction,”IEEE Trans. Cogn. Commun. Netw., vol. 11, no. 2, pp. 738–750, Apr. 2025

    work page 2025

  21. [21]

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

    M. Raissi, P. Perdikaris, and G. E. Karniadakis, “Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,”J. Comput. Phys., vol. 378, pp. 686–707, 2019

    work page 2019

  22. [22]

    PINN and GNN- based RF map construction for wireless communication systems,

    L. Liu, X. Chen, Z. Tang, M. Ma, and W. Zhang, “PINN and GNN- based RF map construction for wireless communication systems,”arXiv preprint arXiv:2507.22513, 2025

    work page arXiv 2025

  23. [23]

    Understanding and mitigating gradient flow pathologies in physics-informed neural networks,

    S. Wang, Y . Teng, and P. Perdikaris, “Understanding and mitigating gradient flow pathologies in physics-informed neural networks,”SIAM J. Sci. Comput., vol. 43, no. 5, pp. A3055–A3081, 2021

    work page 2021

  24. [24]

    From data-driven learning to physics-inspired inferring: A novel mobile MIMO channel prediction scheme based on neural ODE,

    Z. Xiao, Z. Zhang, Z. Chen, Z. Yang, C. Huang, and X. Chen, “From data-driven learning to physics-inspired inferring: A novel mobile MIMO channel prediction scheme based on neural ODE,”IEEE Trans. Wireless Commun., vol. 23, no. 7, pp. 7186–7199, Jul. 2024

    work page 2024

  25. [25]

    Fourier neural operator for parametric partial differential equations,

    Z. Li, N. Kovachki, K. Azizzadenesheli, B. Liu, K. Bhattacharya, A. Stuart, and A. Anandkumar, “Fourier neural operator for parametric partial differential equations,” inProc. Int. Conf. Learn. Represent. (ICLR), 2021

    work page 2021

  26. [26]

    Factorized Fourier neural operators,

    A. Tran, A. Mathews, L. Xie, and C. S. Ong, “Factorized Fourier neural operators,” inProc. Int. Conf. Learn. Represent. (ICLR), 2023

    work page 2023

  27. [27]

    Geometry-informed neural operator for large-scale 3D PDEs,

    Z. Li, N. Kovachki, C. Choy, B. Li, J. Kossaifi, S. Otta, M. A. Nabian, M. Stadler, C. Hundt, K. Azizzadenesheli, and A. Anandkumar, “Geometry-informed neural operator for large-scale 3D PDEs,” inProc. Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 36, 2023, pp. 35 836– 35 854

    work page 2023

  28. [28]

    Learning function-to-function mappings: A Fourier neural operator for next-generation MIMO systems,

    J. Xiao, J. Wang, Q. Sun, Q. Cui, X. Li, D. Niyato, and C.-L. I, “Learning function-to-function mappings: A Fourier neural operator for next-generation MIMO systems,”arXiv preprint arXiv:2510.04664, 2025

    work page arXiv 2025

  29. [29]

    Channel extrapolation for MIMO systems with the assistance of multi-path information induced from channel state information,

    Y . Gao, X. Wu, J. Jiang, Z. Zhang, Z. Yang, S. Xu, C.-X. Wang, and Z. Han, “Channel extrapolation for MIMO systems with the assistance of multi-path information induced from channel state information,”arXiv preprint arXiv:2601.21524, 2026

    work page arXiv 2026

  30. [30]

    Implicit neural representations with periodic activation functions,

    V . Sitzmann, J. Martel, A. Bergman, D. Lindell, and G. Wetzstein, “Implicit neural representations with periodic activation functions,” in Proc. Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 33, 2020, pp. 7462– 7473

    work page 2020

  31. [31]

    NeRF: Representing scenes as neural radiance fields for view synthesis,

    B. Mildenhall, P. P. Srinivasan, M. Tancik, J. T. Barron, R. Ramamoorthi, and R. Ng, “NeRF: Representing scenes as neural radiance fields for view synthesis,” inProc. Eur. Conf. Comput. Vis. (ECCV), ser. Lecture Notes in Computer Science, vol. 12346. Springer, 2020, pp. 405–421

    work page 2020

  32. [32]

    Bayesian active learning for sample efficient 5G radio map reconstruction,

    K. D. Polyzos, A. Sadeghi, W. Ye, S. Sleder, K. Houssou, J. Calder, Z.-L. Zhang, and G. B. Giannakis, “Bayesian active learning for sample efficient 5G radio map reconstruction,”IEEE Trans. Wireless Commun., vol. 23, no. 12, pp. 19 382–19 396, Dec. 2024

    work page 2024

  33. [33]

    T. S. Rappaport,Wireless Communications: Principles and Practice, 2nd ed. Upper Saddle River, NJ, USA: Prentice Hall PTR, 2002

    work page 2002

  34. [34]

    Correlation model for shadow fading in mobile radio systems,

    M. Gudmundson, “Correlation model for shadow fading in mobile radio systems,”Electron. Lett., vol. 27, no. 23, pp. 2145–2146, Nov. 1991

    work page 1991

  35. [35]

    M. F. Levy,Parabolic Equation Methods for Electromagnetic Wave Propagation, ser. IEE Electromagnetic Waves Series. London, U.K.: Institution of Electrical Engineers, 2000, vol. 45

    work page 2000

  36. [36]

    The split-step-fourier and finite-element- based parabolic-equation propagation-prediction tools: Canonical tests, systematic comparisons, and calibration,

    G. Apaydin and L. Sevgi, “The split-step-fourier and finite-element- based parabolic-equation propagation-prediction tools: Canonical tests, systematic comparisons, and calibration,”IEEE Antennas Propag. Mag., vol. 52, no. 3, pp. 66–79, Jun. 2010

    work page 2010

  37. [37]

    FiLM: Visual reasoning with a general conditioning layer,

    E. Perez, F. Strub, H. de Vries, V . Dumoulin, and A. Courville, “FiLM: Visual reasoning with a general conditioning layer,” inProc. AAAI Conf. Artif. Intell., vol. 32, no. 1, 2018, pp. 3942–3951

    work page 2018

  38. [38]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” inProc. Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 30, 2017

    work page 2017

  39. [39]

    Do transformers really perform badly for graph representation?

    C. Ying, T. Cai, S. Luo, S. Zheng, G. Ke, D. He, Y . Shen, and T.-Y . Liu, “Do transformers really perform badly for graph representation?” inProc. Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 34, 2021, pp. 28 877–28 888

    work page 2021

  40. [40]

    Image inpainting for irregular holes using partial convolutions,

    G. Liu, F. A. Reda, K. J. Shih, T.-C. Wang, A. Tao, and B. Catanzaro, “Image inpainting for irregular holes using partial convolutions,” inProc. Eur. Conf. Comput. Vis. (ECCV), ser. Lecture Notes in Computer Science, vol. 11215. Springer, 2018, pp. 89–105

    work page 2018

  41. [41]

    Fourier features let networks learn high frequency functions in low dimensional domains,

    M. Tancik, P. Srinivasan, B. Mildenhall, S. Fridovich-Keil, N. Raghavan, U. Singhal, R. Ramamoorthi, J. Barron, and R. Ng, “Fourier features let networks learn high frequency functions in low dimensional domains,” inProc. Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 33, 2020, pp. 7537–7547

    work page 2020

  42. [42]

    Multi-task learning using uncertainty to weigh losses for scene geometry and semantics,

    A. Kendall, Y . Gal, and R. Cipolla, “Multi-task learning using uncertainty to weigh losses for scene geometry and semantics,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2018, pp. 7482–7491

    work page 2018

  43. [43]

    DeepMIMO: A Generic Deep Learning Dataset for Millimeter Wave and Massive MIMO Applications

    A. Alkhateeb, “DeepMIMO: A generic deep learning dataset for millimeter wave and massive MIMO applications,”arXiv preprint arXiv:1902.06435, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1902