Pith. sign in

REVIEW 2 major objections 4 minor 36 references

A rotatable antenna array on a UAV, jointly optimized with beamforming, raises multiuser sum-rate under QoS limits far above a fixed array.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-10 07:42 UTC pith:TLV4553P

load-bearing objection Solid systems paper: QoS-constrained RAA orientation+beamforming with a careful PDD solver and a practical two-module GNN that actually addresses the gradient-scale problem; gains over FAA are clear under the stated LoS model. the 2 major comments →

arxiv 2607.08420 v1 pith:TLV4553P submitted 2026-07-09 cs.IT math.IT

Optimization and Deep Learning based Resource Allocation for UAV-Aided Wireless Communication with Rotatable Antenna Array

Fengcheng Pei , Lin Xiang , Anja Klein , Robert Schober This is my paper
classification cs.IT math.IT
keywords rotatable antenna arrayUAV communicationsjoint orientation and beamformingpenalty dual decompositiongraph neural networksum-rate maximizationQoS constraints
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper shows that mounting a rotatable linear array of dipole antennas on a hovering UAV, and jointly choosing the array's 3D orientation together with the multiuser beamformers, substantially raises sum-rate while meeting each user's minimum-rate requirement. Unlike a fixed array, the rotatable array can point its high-gain lobe at the users and separate them in angle so that their channels become less correlated. Because orientation and beamforming are tightly coupled, the authors first give a penalty-dual-decomposition algorithm that reaches a KKT point and reliably satisfies tight QoS, then a two-module graph-neural-network that produces nearly the same sum-rate in a few milliseconds and is less sensitive to position errors. Simulations confirm that the rotatable design keeps high feasibility and sum-rate even when the number of antennas is small or the number of users is large—precisely the regimes that matter for size- and power-constrained UAV platforms.

Core claim

Jointly designing the three-dimensional orientation of a rotatable antenna array and the multiuser beamforming vectors yields significantly higher sum-rate and higher probability of meeting per-user rate constraints than the same design performed with a fixed antenna array, under pure line-of-sight channels whose geometry is known to the UAV.

What carries the argument

Penalty dual decomposition (PDD) that introduces auxiliary variables for the effective channel products and the orientation orthogonality constraint, then alternates closed-form beamforming, majorization-minimization SINR updates, and Riemannian conjugate-gradient orientation updates; and a two-module GNN that first predicts orientation (pre-trained on channel-correlation loss) and then beamforming, trained unsupervised with a Lagrangian QoS penalty.

Load-bearing premise

Every user is assumed to have a pure line-of-sight link whose gain and phase are completely determined by known geometry, and the UAV is assumed to know those positions accurately enough to compute orientation and beams.

What would settle it

In a controlled outdoor UAV trial with measured multipath or deliberate localization error of a few meters, check whether the rotatable-array sum-rate and QoS feasibility still exceed those of an identical fixed array by the margins reported in the paper's Figures 6–9.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. The paper equips a hovering multi-antenna UAV with a rotatable antenna array (RAA) whose 3D orientation can be adjusted by a gimbal. It jointly optimizes RAA orientation vectors (ra, rd) and the beamforming matrix W to maximize multiuser sum-rate subject to a total power constraint and per-user minimum-rate (QoS) constraints. The nonconvex problem is first reformulated with auxiliary variables X = H^H W and solved by a double-loop PDD algorithm that yields closed-form updates for W and X (Theorems 1–2) and Riemannian conjugate-gradient steps for the orientations on the unit sphere; the algorithm is claimed to reach a KKT point under mild conditions. A complementary unsupervised GNN framework with two serially connected modules (orientation then beamforming), graph-level pooling, Gram–Schmidt orthogonalization, and a two-stage pre-training/joint-training strategy is introduced to mitigate gradient imbalance and enable fast inference. Extensive simulations under pure LoS geometry compare RAA versus fixed arrays (dipole/isotropic) and optimization versus DL, showing large gains in sum-rate and feasibility, complementary reliability-versus-speed trade-offs, and DL robustness to bounded position errors.

Significance. If the results hold under the stated model, the work supplies a concrete, SWAP-friendly mechanism (simple gimbal rotation of a rigid array) that simultaneously improves multiuser interference management and antenna-element gain alignment for UAV platforms. The dual algorithmic contribution—PDD with closed-form subproblems plus a carefully engineered two-module GNN that preserves permutation equivariance—is technically solid and directly addresses the strong coupling that defeats naïve BCD. The complementary strengths (PDD reliability under stringent QoS, GNN speed and robustness) together with the systematic ablation over N, K, rate thresholds and position-error radius make the paper useful for both offline design and real-time UAV resource allocation. Prior RAA literature is extended by the fairness-aware sum-rate objective rather than pure SE maximization or power minimization.

major comments (2)
  1. [II-A / Remark 1 / Fig. 9] Section II-A, Eqs. (1)–(4) and Remark 1: The entire performance claim rests on pure LoS channels whose steering vectors and element factors are completely determined by known geometry. While bounded position error is examined in Fig. 9, no sensitivity to Rician K-factor, multipath clusters or blockage is provided. Because the claimed RAA gains (high-gain alignment and angular decorrelation) vanish or reverse under rich scattering, a short numerical study with a Rician or geometry-based stochastic model would make the central claim more robust.
  2. [V-B / Fig. 6] Section V-B and the note preceding Fig. 6: Setting the sum-rate of every infeasible realization to zero is transparent for feasibility comparison, yet for high Rk the large Opt–DL gap is driven almost entirely by the feasibility-ratio difference rather than by the rates of the feasible instances. Explicitly reporting (or plotting) the conditional average sum-rate over only the feasible realizations would cleanly separate the two effects and avoid overstating the performance gap.
minor comments (4)
  1. [III-B / Theorem 1] Proof of Theorem 1 is omitted “due to limited space.” A short sketch or an explicit pointer to the identical steps in the cited reference would improve reproducibility.
  2. [Table III] Table III reports wall-clock times on a specific workstation; stating the number of Monte-Carlo trials and whether GPU acceleration was used for the GNN inference would make the speed-up claim fully reproducible.
  3. [III-C / Algorithm 2] Notation for the dual variables and penalty parameters is dense; a short table summarizing the PDD schedule (initial ρ, reduction factors c1/c2, tolerances) would help readers implement Algorithm 2.
  4. [throughout] A few typographical inconsistencies appear (e.g., “UA V” vs. “UAV”, occasional missing spaces around math operators). A careful proof-reading pass is recommended.

Circularity Check

0 steps flagged

No significant circularity: objective/constraints are independent of the solvers; PDD subproblem solutions and unsupervised GNN losses do not redefine the metric as a fit or self-citation.

full rationale

The paper formulates sum-rate maximization under power, orthonormal orientation, and per-user SINR/QoS constraints (P1/C1–C6) from a standard LoS dipole channel model (1)–(4) that is fixed a priori by geometry. The PDD reformulation (fP1)–(P2) introduces auxiliary variables and AL penalties solely to decouple the same constraints; closed-form updates for W (11), X (Theorems 1–2 via MM/KKT), and manifold RCG for ra/rd (17)–(20) are standard first-order/KKT manipulations of those subproblems and do not redefine the objective. The GNN architecture (OM+BM, message passing (23)–(32), Gram–Schmidt/normalization) and two-stage unsupervised loss (P10)–(P11) with LDM duals likewise optimize the original sum-rate + ReLU QoS penalty; no labels or fitted parameters are re-used as “predictions.” Prior RAA citations [11]–[16] (overlapping authors) address distinct objectives (power, SE, EE, completion time, ISAC) and supply only motivational context, not load-bearing uniqueness or ansatz that forces the present sum-rate claim. Simulation comparisons (Figs. 5–9, Table III) evaluate the solvers on independent random user realizations under the stated model. No step reduces a claimed prediction or first-principles result to its own inputs by construction.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 2 invented entities

Central claims rest on standard wireless LoS geometry, dipole array factor models, convex/manifold optimization tools, and GNN message-passing—not on new physical entities. Free parameters are algorithmic and simulation knobs (penalties, learning rates, network widths, QoS level, geometry). Invented structure is the two-module GNN pipeline and the specific PDD splitting for this problem, not a new particle or force.

free parameters (4)
  • PDD penalty/dual schedule (ρ1, ρ2, c1, c2, ε tolerances, Tmax)
    Hand-chosen outer-loop schedule that controls feasibility enforcement and reported convergence; not derived from first principles.
  • GNN architecture widths, attention heads, L1/L2 depth, learning rate, μ step size, E1/E2 epochs
    Table I and §V training settings chosen for performance; affect sum-rate and feasibility of RAA-Dip-DL.
  • Simulation scenario (N=16, K=4, Rk=4 nat/s/Hz, P=1 W, hover height 40 m, 100×100 m area)
    Default operating point for main claims; sweeps vary one axis at a time but claims are conditioned on this LoS geometry class.
  • Path-loss constants in βk (140.7 + 36.7 log10(Dk/1000))
    Taken from cited model [24]; fixed numbers that set absolute SNRs and thus absolute sum-rates.
axioms (5)
  • domain assumption UAV–user channels are pure LoS with known geometry-determined steering vectors and dipole element factors (Eqs. (1)–(4)).
    Stated in §II-A; entire orientation benefit and both solvers depend on this channel structure.
  • domain assumption RAA orientation is fully described by two orthonormal unit vectors ra, rd with free 3D rotation (C2–C4).
    Gimbal idealization; no mechanical limits, rotation energy, or dynamics in (P1).
  • standard math PDD double-loop converges to a KKT point of (P1)/(P2) when constraint violations are driven small enough ([17]–[19]).
    Invoked in §III-D; local optimality only, not global.
  • standard math Strong duality / KKT sufficiency for the W and per-user X quadratic subproblems (Slater, [29]).
    Used for closed forms (11) and Theorem 2.
  • ad hoc to paper Unsupervised Lagrangian penalty on ReLU(γk−γk) plus two-stage training adequately enforces QoS for the GNN.
    §IV-D; soft constraint, not hard feasibility guarantee—explains lower feasibility ratio vs PDD under tight Rk.
invented entities (2)
  • Two-module GNN (OM then BM) with graph-level orientation head, Gram–Schmidt, and two-stage pretrain/joint train no independent evidence
    purpose: Separate orientation and beamforming to mitigate gradient scale imbalance and preserve permutation invariance/equivariance.
    Architecture and training strategy introduced for this paper; independent_evidence false beyond simulation ablations in Fig. 4.
  • PDD splitting of joint RAA orientation–beamforming with auxiliary X and AL penalties on C4/C8 no independent evidence
    purpose: Decouple nonconvex equality coupling so BCD subproblems admit closed forms / RCG.
    Problem-specific application of PDD; not a new physical object.

pith-pipeline@v1.1.0-grok45 · 30170 in / 3564 out tokens · 40806 ms · 2026-07-10T07:42:59.190954+00:00 · methodology

0 comments
read the original abstract

Multi-antenna unmanned aerial vehicle (UAV)-aided communication presents a promising solution to increase the system capacity and improve the quality of service (QoS) of the future wireless networks. In this paper, we equip a UAV platform with a rotatable antenna array (RAA), which can be rotated flexibly in three-dimensional (3D) space via an onboard gimbal, enabling additional spatial degrees of freedom (DoFs) for improving multiuser signal transmission and interference management. Compared with a conventional fixed antenna array (FAA), the RAA can proactively align users with the high-gain region of its antenna elements and reduce the spatial channel correlations among users. To demonstrate the advantages of RAA, we jointly design the RAA orientation and beamforming to maximize the sum-rate of multiple users subject to per-user QoS constraints. The formulated problem is highly nonconvex and exhibits strong coupling between the RAA orientation and beamforming variables. To solve this challenging problem, we propose first an optimization framework based on the penalty dual decomposition (PDD) method to iteratively optimize RAA orientation and beamforming. While the optimization framework yields high reliability in QoS satisfaction and favorable sum-rate performance, its iterative nature may hinder real-time deployment. To accelerate the joint design and preserve a high-quality solution, we further propose a deep learning (DL) framework based on graph neural networks (GNNs). Simulation results demonstrate that RAAs significantly outperform FAAs in UAV-aided communication. Additionally, the proposed optimization framework is capable of satisfying stringent QoS requirements with high reliability, while the proposed DL framework attains comparable sum-rate performance with substantially reduced computation time and exhibits robustness to user position information errors.

Figures

Figures reproduced from arXiv: 2607.08420 by Anja Klein, Fengcheng Pei, Lin Xiang, Robert Schober.

Figure 1
Figure 1. Figure 1: UAV-aided communication system equipped with an RAA. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The architecture of the proposed GNN-based DL framework for joint RAA orientation and beamforming design. [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: (a) Objective value and (b) constraint violation of problem (P2) versus [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: (a) Pre-training loss and (b) joint training loss of Algorithm 3. [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Transmit beampattern gains of schemes (a) FAA-Dip-Opt and (b) [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: (a) Sum-rate and (b) feasibility ratio versus the number of antennas. [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: (a) Sum-rate and (b) feasibility ratio versus the number of users. [PITH_FULL_IMAGE:figures/full_fig_p012_8.png] view at source ↗

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

36 extracted references · 36 canonical work pages

  1. [1]

    A Survey on Millimeter-Wave Beamforming Enabled UA V Communications and Networking,

    Z. Xiao et al., “A Survey on Millimeter-Wave Beamforming Enabled UA V Communications and Networking,”IEEE Commun. Surveys Tuts., vol. 24, no. 1, pp. 557-610, 2022

    work page 2022

  2. [2]

    A Tutorial on UA Vs for Wireless Networks: Applications, Challenges, and Open Problems,

    M. Mozaffari et al., “A Tutorial on UA Vs for Wireless Networks: Applications, Challenges, and Open Problems,”IEEE Commun. Surveys Tuts., vol. 21, no. 3, pp. 2334-2360, 2019

    work page 2019

  3. [3]

    Noncooperative Cellular Wireless with Unlimited Num- bers of Base Station Antennas,

    T. L. Marzetta, “Noncooperative Cellular Wireless with Unlimited Num- bers of Base Station Antennas,”IEEE Trans. Wireless Commun., vol. 9, no. 11, pp. 3590-3600, Nov. 2010

    work page 2010

  4. [4]

    A Comprehensive Overview on 5G-and-beyond Networks with UA Vs,

    Q. Wu, et al., “A Comprehensive Overview on 5G-and-beyond Networks with UA Vs,”IEEE J. Sel. Areas Commun., vol. 39, no. 10, pp. 2912-2945, Oct. 2021

    work page 2021

  5. [5]

    Joint Maneuver and Beamforming Design for UA V-enabled Integrated Sensing and Communication,

    Z. Lyu, G. Zhu and J. Xu, “Joint Maneuver and Beamforming Design for UA V-enabled Integrated Sensing and Communication,”IEEE Trans. Wireless Commun., vol. 22, no. 4, pp. 2424-2440, Apr. 2023

    work page 2023

  6. [6]

    Joint Beamforming and Trajectory Optimization for Multi-UA V-Assisted Integrated Sensing and Communication Systems,

    Y . K. Tun et al., “Joint Beamforming and Trajectory Optimization for Multi-UA V-Assisted Integrated Sensing and Communication Systems,” IEEE Trans. Veh. Technol

    work page

  7. [7]

    Rotatable Antenna Enabled Wireless Communication and Sensing: Opportunities and Challenges,

    B. Zheng et al., “Rotatable Antenna Enabled Wireless Communication and Sensing: Opportunities and Challenges,”IEEE Wireless Commun., vol. 33, no. 3, pp. 134-141, Jun. 2026

    work page 2026

  8. [8]

    Movable Antennas for Wireless Communication: Opportu- nities and Challenges,

    L. Zhu et al., “Movable Antennas for Wireless Communication: Opportu- nities and Challenges,”IEEE Commun. Mag., vol. 62, no. 6, pp. 114-120, June 2024

    work page 2024

  9. [9]

    Rotatable Antenna-Enabled Wireless Communication: Modeling and Optimization,

    B. Zheng et al., “Rotatable Antenna-Enabled Wireless Communication: Modeling and Optimization,”IEEE Trans. Wireless Commun., vol. 74, pp. 6825-6842, 2026

    work page 2026

  10. [10]

    6D Movable Antenna Based on User Distribution: Modeling and Optimization,

    X. Shao et al., “6D Movable Antenna Based on User Distribution: Modeling and Optimization,”IEEE Trans. Wireless Commun., vol. 24, no. 1, pp. 355-370, Jan. 2025

    work page 2025

  11. [11]

    Rotatable Antenna Array Enabled UA V mmWave Massive MIMO Communication,

    X. Zhang et al., “Rotatable Antenna Array Enabled UA V mmWave Massive MIMO Communication,”IEEE Trans. Wireless Commun., vol. 74, pp. 1219-1236, 2026

    work page 2026

  12. [12]

    Joint Optimization of Beamforming and 3D Array Steering for Multi-Antenna UA V Communications,

    L. Xiang et al., “Joint Optimization of Beamforming and 3D Array Steering for Multi-Antenna UA V Communications,”IEEE WCNC, Dubai, 2024

    work page 2024

  13. [13]

    Energy-Efficient Dynamic Array-Steering and Beam- forming for UA V-Aided Communications

    L. Xiang et al., “Energy-Efficient Dynamic Array-Steering and Beam- forming for UA V-Aided Communications”,IEEE GlobeCom, Cape Town, 2024

    work page 2024

  14. [14]

    Completion Time Minimization for UA V-aided Com- munications with Rotatable Dipole Array

    B. Yilmaz et al., “Completion Time Minimization for UA V-aided Com- munications with Rotatable Dipole Array”,IEEE ICC, Montreal, 2025

    work page 2025

  15. [15]

    Joint Optimization of Beamforming and 3D Array-Steering for UA V-Aided ISAC,

    F. Pei et al., “Joint Optimization of Beamforming and 3D Array-Steering for UA V-Aided ISAC,”IEEE ICC, Denver, CO, USA, 2024

    work page 2024

  16. [16]

    Transmit Beamforming and Array Steering Optimization for UA V-Aided Bistatic ISAC

    F. Pei et al., “Transmit Beamforming and Array Steering Optimization for UA V-Aided Bistatic ISAC”,IEEE GlobeCom, Cape Town, 2024

    work page 2024

  17. [17]

    Penalty Dual Decomposition Method for Nonsmooth Nonconvex Optimization—Part I: Algorithms and Convergence Analy- sis,

    Q. Shi et al., “Penalty Dual Decomposition Method for Nonsmooth Nonconvex Optimization—Part I: Algorithms and Convergence Analy- sis,”IEEE Trans. Signal Process., vol. 68, pp. 4108-4122, 2020. 14

    work page 2020

  18. [18]

    Penalty Dual Decomposition Method for Nonsmooth Nonconvex Optimization—Part II: Applications,

    Q. Shi et al., “Penalty Dual Decomposition Method for Nonsmooth Nonconvex Optimization—Part II: Applications,”IEEE Trans. Signal Process., vol. 68, pp. 4242-4257, 2020

    work page 2020

  19. [19]

    Spectral Efficiency Optimization For Millimeter Wave Multiuser MIMO Systems,

    Q. Shi et al., “Spectral Efficiency Optimization For Millimeter Wave Multiuser MIMO Systems,”IEEE J. Select. Topics Signal Process., vol. 12, no. 3, pp. 455-468, Jun. 2018

    work page 2018

  20. [20]

    Graph Neural Networks for Scalable Radio Resource Management: Architecture Design and Theoretical Analysis,

    Y . Shen et al., “Graph Neural Networks for Scalable Radio Resource Management: Architecture Design and Theoretical Analysis,”IEEE J. Select. Areas Commun., vol. 39, no. 1, pp. 101-115, Jan. 2021

    work page 2021

  21. [21]

    Graph Neural Networks for Wireless Communications: From Theory to Practice,

    Y . Shen et al., “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

  22. [22]

    GNN-Based Beamforming for Sum-Rate Maximization in MU-MISO Networks,

    Y . Li et al., “GNN-Based Beamforming for Sum-Rate Maximization in MU-MISO Networks,”IEEE Trans. Wireless Commun., vol. 23, no. 8, pp. 9251-9264, Aug. 2024

    work page 2024

  23. [23]

    DeepSIC: Deep Soft Interference Cancellation for Multiuser MIMO Detection,

    N. Shlezinger et al., “DeepSIC: Deep Soft Interference Cancellation for Multiuser MIMO Detection,”IEEE Trans. Wireless Commun., vol. 20, no. 2, pp. 1349-1362, Feb. 2021

    work page 2021

  24. [24]

    On Stream Selection for Interference Alignment in Heterogeneous Networks,

    A. Beyazıt et al., “On Stream Selection for Interference Alignment in Heterogeneous Networks,”J. Wireless Commun. Network, 80 (2016)

    work page 2016

  25. [25]

    C. A. Balanis,Antenna Theory: Analysis and Design. 4th ed, 2016

    work page 2016

  26. [26]

    Multiuser MISO UA V Communications in Uncertain Environments With No-Fly Zones: Robust Trajectory and Resource Allocation Design,

    D. Xu et al., “Multiuser MISO UA V Communications in Uncertain Environments With No-Fly Zones: Robust Trajectory and Resource Allocation Design,”IEEE Trans. Commun., vol. 68, no. 5, pp. 3153-3172, May 2020

    work page 2020

  27. [27]

    Energy-Efficient UA V Communications in the Presence of Wind: 3D Modeling and Trajectory Design,

    X. Dai et al., “Energy-Efficient UA V Communications in the Presence of Wind: 3D Modeling and Trajectory Design,”IEEE Trans. Wireless Commun., vol. 23, no. 3, pp. 1840-1854, March 2024

    work page 2024

  28. [28]

    Multiplier and Gradient Methods,

    M.R.Hestenes, “Multiplier and Gradient Methods,”J. Optim. Theory Appl., vol. 4, no. 5, pp. 303–320, 1969

    work page 1969

  29. [29]

    Boyd et al.,Convex Optimization, Cambridge University Press, 2004

    S. Boyd et al.,Convex Optimization, Cambridge University Press, 2004

    work page 2004

  30. [30]

    Massive MIMO Multicasting With Finite Blocklength,

    X. Zhang et al., “Massive MIMO Multicasting With Finite Blocklength,” IEEE Trans. Wireless Commun., vol. 23, no. 10, pp. 15018-15034, Oct. 2024

    work page 2024

  31. [31]

    Majorization-Minimization Algorithms in Signal Pro- cessing, Communications, and Machine Learning,

    Y . Sun et al., “Majorization-Minimization Algorithms in Signal Pro- cessing, Communications, and Machine Learning,”IEEE Trans. Signal Process., vol. 65, no. 3, pp. 794-816, Feb.1, 2017

    work page 2017

  32. [32]

    Absil et al.,Optimization Algorithms on Matrix Manifolds.Prince- ton University Press, 2009

    P.-A. Absil et al.,Optimization Algorithms on Matrix Manifolds.Prince- ton University Press, 2009

    work page 2009

  33. [33]

    Nocedal et al.,Numerical Optimization, 2nd Edition, Springer, 2006

    J. Nocedal et al.,Numerical Optimization, 2nd Edition, Springer, 2006

    work page 2006

  34. [34]

    Weighted Sum-Rate Maximization for MISO Down- link Cellular Networks via Branch and Bound,

    S. K. Joshi et al., “Weighted Sum-Rate Maximization for MISO Down- link Cellular Networks via Branch and Bound,”IEEE Trans. Signal Process., vol. 60, no. 4, pp. 2090-2095, April 2012

    work page 2090

  35. [35]

    W. L. Hamilton,Graph Representation Learning, Morgan & Claypool, 2020

    work page 2020

  36. [36]

    Adaptive Beam-Frequency Allocation Algorithm with Position Uncertainty for Millimeter-Wave MIMO Systems,

    R. Ismayilov et al., “Adaptive Beam-Frequency Allocation Algorithm with Position Uncertainty for Millimeter-Wave MIMO Systems,”IEEE VTC Spring, Porto, Portugal, 2018

    work page 2018