Signal Processing · arXiv:2611.16305 v1 · submitted 12 Nov 2026 · 44 references
Physics-Aware Transformer-Based Beamforming Optimization for Near-Field E-MIMO ISAC Systems
The paper says
“This paper proposes a physics-aware Transformer-based beamforming optimization (PATBO) framework for near-field extreme massive multiple-input multiple-output integrated sensing and communication (E-MIMO-ISAC) systems. PATBO combines regularized zero-forcing (RZF) communication directions and spherical-wave near-field sensing directions with learned stream-power allocation.”
more of the abstract
A token-preserving Transformer maintains the correspondence between individual users/targets and their allocated powers, while a Cramér–Rao bound (CRB)-sensitivity-aware objective emphasizes power-allocation errors with greater localization impact. Physical feasibility is explicitly verified after inference, with infeasible predictions handled through physics-constrained power repair and optimizer fallback. For a 32 × 32 E-MIMO array with 1024 antennas, the proposed structured representation reduces the learned output from 12,288 real-valued coefficients of an unstructured complex beamformer to six stream-power variables. Direct PATBO achieves 100.98% reference-rate retention and 19.54% joint feasibility, which increases to 87.79% after repair. The complete hybrid framework achieves 99.89% reference-rate retention and 100% constraint satisfaction on the retained test set, while requiring optimizer fallback for only 12.21% of the scenarios. Direct PATBO inference requires 1.11 ms per sample, corresponding to approximately 858 × and 779 × lower runtime than the evaluated cold-start and PATBO-warm-start sequential quadratic programming (SQP) implementations, respectively.
from the abstract
Pith says
Machine review · deepseek-v4-flash · 13 Nov 2026 · what was read was not recorded · not checked by a person
It can be wrong. Challenge this review or reply to it.
A transparent learning-to-optimize paper whose real contribution is the repair/fallback cascade; the headline speedup applies only to the 19.5% direct-feasible path, and the paper never reports end-to-end hybrid latency. This line is wrong
the referee pass did not run
How often objections from reviews like this one are right: not yet measured
Not checked: figures, code, what the 44 references say.
What would settle it
Measure end-to-end wall-clock latency of the complete PATBO pipeline on the 696 retained test samples, from scenario input to the final beamformer selected by (51), using the actual physics-repair routine for all directly infeasible predictions and an actual optimizer fallback for the post-repair failures, then compare the mean to cold-start and warm-start SQP on the same samples. If the end-to-end mean is not below the SQP baselines, the 858x/779x claims should be re-scoped to direct mode only and the paper should report the per-mode latency and feasibility tradeoff explicitly.
from the skeptic's reading
How the review went
3 readings by deepseek-v4-flash · 13 Nov 2026
- Reader holds up
- Skeptic partly agrees: says it holds up only with conditions The 858x speedup is reported only for the direct inference path, but that path is feasible on just 19.54% of test scenarios; weighting the reported repair and fallback latencies by the reported operating-mode distribution gives an estimated ~1.74 s full-pipeline average, so the low-latency claim…
- Editor wrote the verdict above
Not run on this paper: the referee, the simulated authors' defence.
Pith did not check this review's objections against the paper.
The review in full
The review in brief, machine-written
PATBO is a serious, honest engineering paper. It reduces near-field E-MIMO-ISAC beamforming to six learned stream powers on fixed RZF and near-field focusing directions, then spends most of its effort on what happens when the learned allocation is infeasible. The verify-repair-fallback cascade and the CRB-sensitivity weighting are the real contributions, and the paper reports its failure modes plainly: direct joint feasibility is only 19.54%, repair succeeds on 84.82% of the attempts, 12.21% of scenes fall back, and the 100% feasibility figure is explicitly empirical on the retained test set. That transparency earns trust.
The system model, CRB setup, and constraint checks are standard and clearly stated. The arithmetic is consistent: 136 directly feasible plus 475 repaired plus 85 fallback equals 696 test samples. The ablation against an MLP and pooled Transformer variants supports the token-preserving readout, and the 2048x output-dimension reduction relative to unstructured beamformer prediction is a solid argument for the structured representation. I also credit the authors for stating that the SQP baseline is a local solver with only 39–50% feasibility. The citation pattern looks appropriate for the subfield.
Two soft spots. The significant one is the latency claim. The 858x speedup applies to the direct path alone, and that path is jointly feasible on only 19.54% of retained test scenes. The paper never reports end-to-end hybrid latency. Weighting their own Table V numbers by the operating-mode distribution, with warm-start SQP as a stand-in for online fallback, gives roughly 1.74 s per sample: about two times the warm-start SQP time, because 68% of scenes go through the 2.4 s repair stage. The comparison in the abstract is therefore between a fast but mostly infeasible predictor and slower numerical solvers, not between two feasible end-to-end systems. This is fixable in revision with one paragraph and a small benchmark, but as written it is the main gap in the paper's value proposition. The second soft spot is the fixed RZF/focusing subspace: both the reference labels and the learned predictions live in that subspace, so the results do not validate the subspace itself. The authors acknowledge this in Section II-E, so it is a scope limit rather than a flaw.
Who should read it: people designing learned beamformers for near-field ISAC or XL-MIMO systems. The framework is clearly described, the limitations are stated rather than hidden, and the repair/fallback hierarchy is worth borrowing. I would send it to peer review; the end-to-end latency gap should be addressed by the revision, but it does not undermine the core feasibility machinery.
The strongest objection
The central low-latency claim is tied to direct PATBO inference, but direct predictions are jointly feasible on only 19.54% of the retained test set. The 100%-feasible hybrid result is obtained by sending 68.25% of samples through physics-constrained repair and 12.21% through optimizer fallback. Combining the paper's own reported stage latencies (Table V: direct 1.1085 ms, repair 2.4011 s, warm-start SQP 0.8641 s) with the reported operating-mode fractions gives an expected hybrid latency of roughly 1.74 s per sample, i.e., about 1.8x the cold-start SQP time and 2.0x the warm-start SQP time. The paper never reports end-to-end hybrid latency, and the advertised 858x/779x speedups apply only to the direct mode, which does not solve the constrained problem on 80.5% of samples. The comparison is therefore between a fast but mostly infeasible predictor and slower numerical solvers, not between two feasible systems. This missing end-to-end latency estimate is the main unaddressed gap in the paper's low-latency value proposition.
The reader's weakest-assumption concern about the fixed RZF/focusing beam-direction subspace is real but is explicitly acknowledged in Section II-E, and the internal comparison is fair within that subspace, so I do not treat it as the decisive issue. The more consequential gap is that the paper's headline runtime advantage is not measured for the only operating mode that achieves 100% feasibility. From the paper's own Section IV-E and Table V, direct PATBO handles 136/696 samples, repair handles 475/696, and fallback handles 85/696; a weighted average of the reported per-stage runtimes is about 1.74 s per sample. This is an estimate, not a measurement, but it shifts the burden to the authors to report actual end-to-end hybrid latency. The proposed framework may still be valuable as a warm-start or as a direct mode for latency-tolerant soft-feasibility use cases, and the paper is transparent about the empirical nature of the 100% feasibility result. Because the fix is a single measurement plus careful scoping of the speedup claim, conditional acceptance is the appropriate outcome rather than rejection.
Pith's reading of the paper (a machine's summary, not the authors')
This paper proposes PATBO, a physics-aware Transformer framework for near-field extreme massive MIMO integrated sensing and communication (E-MIMO-ISAC) systems. PATBO separates beamforming into two parts: physically constructed beam directions (regularized zero-forcing for communication, spherical-wave focusing for sensing) and a learned stream-power allocation. The Transformer predicts only K+L normalized stream-power fractions (six variables for a 32x32 array with K=4 users, L=2 targets), reducing the learned output from 12,288 real-valued coefficients. After inference, physical feasibility is explicitly verified; infeasible predictions undergo physics-constrained power repair, and optimizer fallback is used only when repair fails. The hybrid framework achieves 99.89% reference-rate retention and 100% constraint satisfaction on the retained test set, with direct inference at 1.11 ms per sample (about 858x faster than cold-start SQP).
What carries the argument. The central mechanism is the structured beamforming representation that separates beam directions from power allocation. Communication beam directions are constructed using regularized zero-forcing (RZF), and sensing beam directions use spherical-wave near-field focusing vectors. The Transformer predicts only the K+L normalized stream-power fractions via a token-preserving architecture with CRB-sensitivity-aware training. The softmax output ensures the unit-simplex constraint, while the total transmit power is structurally enforced. A hierarchical verify-repair-fallback strategy handles constraint satisfaction.
Core claim. The paper establishes that a structured, physics-aware approach to beamforming optimization for near-field E-MIMO-ISAC systems can dramatically reduce the computational burden of learning-based beamforming while maintaining high feasibility and communication/sensing performance. By fixing beam directions to physically meaningful constructions (RZF for communication, spherical-wave focusing for sensing) and learning only the stream-power allocation via a token-preserving Transformer with CRB-sensitivity-aware training, the framework reduces the learned output from 12,288 real-valued coefficients to just six stream-power variables. The hybrid verify-repair-fallback strategy achieves 100% constraint satisfaction on the retained test set, with only 12.21% of scenarios requiring optimizer fallback. Direct inference runs at 1.11 ms per sample, approximately 858x faster than cold-start SQP, and this latency remains nearly constant as the array size grows from M=400 to M=1024.
The premise it rests on. The structured beam directions (RZF for communication, spherical-wave focusing for sensing) are fixed and assumed sufficient for near-optimal performance, and the reference optimizer used for training labels and fallback operates within the same K+L stream-power subspace, so the comparison does not validate the subspace itself.
If the paper is right (Pith's inferences, not the authors')
Near-field E-MIMO-ISAC beamforming can be reduced to a low-dimensional power-allocation problem when physically meaningful beam directions are fixed, enabling practical real-time operation.
The token-preserving Transformer architecture with CRB-sensitivity weighting provides a template for learning in other constrained beamforming problems where different streams have different sensitivities to prediction errors.
The hybrid verify-repair-fallback approach offers a practical tradeoff between latency and feasibility, applicable to other wireless resource allocation problems where direct learning cannot guarantee constraint satisfaction.
The structured representation scales favorably with array size: inference latency remains ~1.1 ms for M=400-1024, making it suitable for future extremely large antenna arrays.
The framework demonstrates that learning-based beamforming can achieve 100% constraint satisfaction on the retained test set while being orders of magnitude faster than numerical optimization, though this guarantee is empirical rather than theoretical.
The structured beam directions (RZF for communication, spherical-wave focusing for sensing) are fixed and assumed sufficient; the paper does not validate this subspace itself. If these directions exclude good beamformers for some geometries, both the reference and learned solutions are suboptimal.
The 100% final feasibility is an empirical result on the retained test set (where the reference optimizer already found a feasible solution). It does not guarantee feasibility for arbitrary unseen scenarios, as the authors acknowledge.
The CRB-sensitivity-aware loss is computed using the reference solution's sensitivity, not the predicted one. This could be extended to iterative sensitivity estimation during training.
The framework could be extended to dynamic scenarios with user/target mobility by incorporating temporal or predictive tokens, as suggested in future work.
Check this review
Pith stored no record of this review's passes and took no fingerprint of it. Reviews made by code from 13 Nov 2026 on carry both.
How often reviews like this one are right
Reviews like this one: written by deepseek-v4-flash, where what was read was not recorded.
How often objections from reviews like this one are right: not yet measured
Reviews by people
This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.
The paper
Abstract
This paper proposes a physics-aware Transformer-based beamforming optimization (PATBO) framework for near-field extreme massive multiple-input multiple-output integrated sensing and communication (E-MIMO-ISAC) systems. PATBO combines regularized zero-forcing (RZF) communication directions and spherical-wave near-field sensing directions with learned stream-power allocation. A token-preserving Transformer maintains the correspondence between individual users/targets and their allocated powers, while a Cramér–Rao bound (CRB)-sensitivity-aware objective emphasizes power-allocation errors with greater localization impact. Physical feasibility is explicitly verified after inference, with infeasible predictions handled through physics-constrained power repair and optimizer fallback. For a 32 × 32 E-MIMO array with 1024 antennas, the proposed structured representation reduces the learned output from 12,288 real-valued coefficients of an unstructured complex beamformer to six stream-power variables. Direct PATBO achieves 100.98% reference-rate retention and 19.54% joint feasibility, which increases to 87.79% after repair. The complete hybrid framework achieves 99.89% reference-rate retention and 100% constraint satisfaction on the retained test set, while requiring optimizer fallback for only 12.21% of the scenarios. Direct PATBO inference requires 1.11 ms per sample, corresponding to approximately 858 × and 779 × lower runtime than the evaluated cold-start and PATBO-warm-start sequential quadratic programming (SQP) implementations, respectively.
The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.
abstract compiled · 2026-11-13
The authors' main claim, as the classifier picked it out 0.97
“The complete hybrid framework achieves $99.89\%$ reference-rate retention and $100\%$ constraint satisfaction on the retained test set, while requiring optimizer fallback for only $12.21\%$ of the scenarios.”
How the main claim is stated: asserted 1.00
What the abstract itself states, as probabilities: claims first 0.05 · states limitation 0.78 · human participants 0.02 · controlled comparison 0.84 · randomized assignment 0.01 · states quantitative result 0.99 · compares with prior results 0.93 · mentions code or data availability 0.02
The abstract, sentence by sentence
This table labels the abstract only (arXiv listing abstract v1). The number is the probability the classifier gave the label.
| Sentence | Role | The authors' own work | States the main conclusion | Strength |
|---|---|---|---|---|
| This paper proposes a physics-aware Transformer-based beamforming optimization (PATBO) framework for near-field extreme massive multiple-input multiple-output integrated sensing and communication (E-MIMO-ISAC) systems. | contribution 0.99 | 0.88 | 0.10 | asserted 1.00 |
| PATBO combines regularized zero-forcing (RZF) communication directions and spherical-wave near-field sensing directions with learned stream-power allocation. | method 0.97 | 0.81 | 0.05 | asserted 1.00 |
| A token-preserving Transformer maintains the correspondence between individual users/targets and their allocated powers, while a Cram\'er--Rao bound (CRB)-sensitivity-aware objective emphasizes power-allocation errors with greater localization impact. | method 0.98 | 0.85 | 0.04 | asserted 1.00 |
| Physical feasibility is explicitly verified after inference, with infeasible predictions handled through physics-constrained power repair and optimizer fallback. | method 0.99 | 0.91 | 0.05 | asserted 1.00 |
| For a $32\times32$ E-MIMO array with $1024$ antennas, the proposed structured representation reduces the learned output from $12{,}288$ real-valued coefficients of an unstructured complex beamformer to six stream-power variables. | method 0.72 | 0.92 | 0.06 | asserted 1.00 |
| Direct PATBO achieves $100.98\%$ reference-rate retention and $19.54\%$ joint feasibility, which increases to $87.79\%$ after repair. | result 1.00 | 0.96 | 0.15 | asserted 1.00 |
| The complete hybrid framework achieves $99.89\%$ reference-rate retention and $100\%$ constraint satisfaction on the retained test set, while requiring optimizer fallback for only $12.21\%$ of the scenarios. | result 1.00 | 0.95 | 0.33 | asserted 1.00 |
| Direct PATBO inference requires $1.11$~ms per sample, corresponding to approximately $858\times$ and $779\times$ lower runtime than the evaluated cold-start and PATBO-warm-start sequential quadratic programming (SQP) implementations, respectively. | result 1.00 | 0.95 | 0.11 | asserted 1.00 |
What kind of paper the classifier reports this is: method or tool 1.00 · evidence from simulation or computation 1.00
What was examined, and what was not
Sources read
- arXiv listing abstract v1
Not examined
- full text
- figures and tables
- references and citing sentences
- relations to other papers
Levels run
- P abstract
Not run
- S sections
- T body sentences
- C citations
- R cross-paper relations
How the labels were made
Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-11-13: 1 request, 6,620 input tokens.
Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON
Figures (6)
Figures are the authors', reproduced from the paper.
References (44)
- [1] F. Liu, Y . Cui, C. Masouros, J. Xu, T. X. Han, Y . C. Eldar, and S. Buzzi, “Integrated sensing and communications: Toward dual-functional wire- less networks for 6G and beyond,”IEEE J. Sel. Areas Commun., vol. 40, no. 6, pp. 1728–1767, Jun. 2022 on Pith
- [2] M. Ali, A. A. Badrudeen, S. Baek, Y . Kim, and S. Kim, “Joint beamforming design under target reflected interference in RIS-assisted bi-static ISAC system,” inProc. Int. Conf. Artif. Intell. Inf. Commun. (ICAIIC), Tokyo, Japan, 2026, pp. 1259–1264. 13 on Pith
- [3] Z. Wang, J. Zhang, H. Du, D. Niyato, S. Cui, B. Ai, M. Debbah, K. B. Letaief, and H. V . Poor, “A tutorial on extremely large-scale MIMO for 6G: Fundamentals, signal processing, and applications,”IEEE Commun. Surveys Tuts., vol. 26, no. 3, pp. 1560–1605, 3rd Quart. 2024 on Pith
- [4] H. Lu, Y . Zeng, C. You, Y . Han, J. Zhang, Z. Wang, Z. Dong, S. Jin, C.-X. Wang, T. Jiang, X. You, and R. Zhang, “A tutorial on near-field XL-MIMO communications toward 6G,”IEEE Commun. Surveys Tuts., vol. 26, no. 4, pp. 2213–2257, 4th Quart. 2024 on Pith
- [5] C. Ouyang, Z. Wang, Y . Chen, X. Mu, and P. Zhu, “A primer on near- field communications for next-generation multiple access,”Proc. IEEE, vol. 112, no. 9, pp. 1527–1565, Sep. 2024 on Pith
- [6] X. Li, Z. Dong, and Y . Zeng, “Multi-user modular XL-MIMO commu- nications: Near-field beam focusing pattern and user grouping,”IEEE Trans. Wireless Commun., vol. 23, no. 10, pp. 13 766–13 781, Oct. 2024 on Pith
- [7] H. Wang, Z. Xiao, and Y . Zeng, “Cram´er–rao bounds for near-field sens- ing with extremely large-scale MIMO,”IEEE Trans. Signal Process., vol. 72, pp. 701–717, 2024 on Pith
- [8] K. Qu, S. Guo, N. Saeed, and J. Ye, “Near-field integrated sensing and communication: Performance analysis and beamforming design,”IEEE Open J. Commun. Soc., vol. 5, pp. 6353–6366, Oct. 2024 on Pith
- [9] A. Ray, S. Majhi, H. Rajput, and M. Di Renzo, “ISAC for near-field and extremely large-scale uniform planar array systems,”IEEE Trans. Commun., vol. 74, pp. 1497–1512, 2026 on Pith
- [10] H. Li, Z. Wang, X. Mu, Z. Pan, and Y . Liu, “Near-field integrated sensing, positioning, and communication: A downlink and uplink frame- work,”IEEE J. Sel. Areas Commun., vol. 42, no. 9, pp. 2196–2212, Sep. 2024 on Pith
- [11] B. Zhao, C. Ouyang, Y . Liu, X. Zhang, and H. V . Poor, “Modeling and analysis of near-field ISAC,”IEEE J. Sel. Topics Signal Process., vol. 18, no. 4, pp. 678–693, May 2024 on Pith
- [12] X. Wang, W. Zhai, X. Wang, M. G. Amin, and K. Cai, “Wideband near-field integrated sensing and communication with sparse transceiver design,”IEEE J. Sel. Topics Signal Process., vol. 18, no. 4, pp. 662–677, May 2024 on Pith
- [13] Y . Lin, Z. Liu, J. Zhang, F. Liu, X. Li, Q. Zhang, Z. Wei, S. Fan, and J. Yan, “Near-field integrated sensing and communication beamforming considering complexity,”IEEE Trans. Veh. Technol., vol. 74, no. 4, pp. 6021–6034, Apr. 2025 on Pith
- [14] C. Meng, D. Ma, Z. Wang, Y . Liu, Z. Wei, and Z. Feng, “Near-field hybrid beamforming design for modular XL-MIMO ISAC systems,” IEEE Trans. Commun., vol. 73, no. 11, pp. 11 840–11 854, Nov. 2025 on Pith
- [15] J. Ding, Z. Zhou, X. Shao, B. Jiao, and R. Zhang, “Movable antenna- aided near-field integrated sensing and communication,”IEEE Trans. Wireless Commun., vol. 25, pp. 493–508, 2026 on Pith
- [16] Y . Kim, S.-H. Park, S. Yang, K.-K. Wong, L. Dai, and C.-B. Chae, “Near-field challenges in ultra-wideband ISAC: Beamforming strategies and system insights,”IEEE Commun. Mag., pp. 1–7, May 2026 on Pith
- [17] H. Lei, J. Zhang, Z. Wang, B. Ai, and E. Bj ¨ornson, “Near-field user localization and channel estimation for XL-MIMO systems: Fundamen- tals, recent advances, and outlooks,”IEEE Wireless Commun., vol. 32, no. 4, pp. 190–198, Aug. 2025 on Pith
- [18] F. Zeinali, S. Javadi, M. R. Mili, M. B. Mashhadi, A. Al-Dweik, and N. Al-Dhahir, “Beamforming design for near-field fim-aided isac systems: A drl approach,”IEEE Trans. Veh. Technol., vol. 75, no. 8, pp. 18 467–18 472, Aug. 2026 on Pith
- [19] G. Jiang and C. Qi, “Near-field beam training based on deep learning for extremely large-scale MIMO,”IEEE Commun. Lett., vol. 27, no. 8, pp. 2063–2067, Aug. 2023 on Pith
- [20] X. Shi, J. Wang, X. Wang, C. You, and J. Song, “Double-sided near-field XL-MIMO: Beamfocusing codeword selection and channel estimation,” IEEE Trans. Commun., vol. 73, no. 5, pp. 3441–3455, May 2025 on Pith
show 24 more
- [21] Q. Yang, A. Guerra, F. Guidi, N. Shlezinger, H. Zhang, L.-X. Yang, D. Dardari, B. Wang, and Y . C. Eldar, “Beam focusing for near-field multi-user localization,”IEEE Trans. Veh. Technol., vol. 74, no. 8, pp. 12 259–12 273, Aug. 2025 on Pith
- [22] Y . Cui, H. Ding, Y . Ma, X. Li, H. Zhang, and Y . Fang, “Energy-efficient integrated sensing and communication in collaborative millimeter wave networks,”IEEE Trans. Wireless Commun., vol. 24, no. 3, pp. 2341– 2357, Mar. 2025 on Pith
- [23] Y . Zhao, Q. Wu, W. Chen, Y . Zeng, R. Liu, W. Mei, F. Hou, and S. Ma, “Multi-functional beamforming design for integrated sensing, communication, and computation,”IEEE Trans. Commun., vol. 73, no. 8, pp. 6322–6336, Aug. 2025 on Pith
- [24] S. Zhang, W. Hao, G. Sun, C. Huang, Z. Zhu, X. Li, and C. Yuen, “Joint beamforming optimization for active STAR-RIS-assisted ISAC systems,” IEEE Trans. Wireless Commun., vol. 23, no. 11, pp. 15 888–15 902, Nov. 2024 on Pith
- [25] H. Hua, J. Xu, and R. Zhang, “Near-field integrated sensing and communication with extremely large-scale antenna array,”IEEE Trans. Wireless Commun., vol. 24, no. 12, pp. 9962–9977, Dec. 2025 on Pith
- [26] M. Sun, Y . Gong, R. Zhang, Y . Zhao, X. Jing, C. Yuen, and D. W. K. Ng, “Robust spherical wavefront beamforming for near-field ISAC with MMSE optimization: A distance-angle perspective,”IEEE Trans. Commun., vol. 74, pp. 6718–6733, 2026 on Pith
- [27] W. Mao, Y . Lu, C.-Y . Chi, B. Ai, Z. Zhong, and Z. Ding, “Communica- tion-sensing region for cell-free massive MIMO ISAC systems,”IEEE Trans. Wireless Commun., vol. 23, no. 9, pp. 12 396–12 411, Sep. 2024 on Pith
- [28] U. Demirhan and A. Alkhateeb, “Cell-free ISAC MIMO systems: Joint sensing and communication beamforming,”IEEE Trans. Commun., vol. 73, no. 6, pp. 4454–4468, Jun. 2025 on Pith
- [29] W. Xia, G. Zheng, Y . Zhu, J. Zhang, J. Wang, and A. P. Petropulu, “A deep learning framework for optimization of MISO downlink beam- forming,”IEEE Trans. Commun., vol. 68, no. 3, pp. 1866–1880, Mar. 2020 on Pith
- [30] J. Zhang, W. Xia, M. You, G. Zheng, S. Lambotharan, and K.-K. Wong, “Deep learning enabled optimization of downlink beamforming under per-antenna power constraints: Algorithms and experimental demonstra- tion,”IEEE Trans. Wireless Commun., vol. 19, no. 6, pp. 3738–3752, Jun. 2020 on Pith
- [31] Y . Li, Y .-F. Liu, F. Xu, Q. Shi, and T.-H. Chang, “Learning to optimize QoS-constrained beamforming in multi-user systems: A penalty-dual framework,”IEEE Trans. Wireless Commun., vol. 23, no. 11, pp. 16 123– 16 138, Nov. 2024 on Pith
- [32] Y . Li and Y .-F. Liu, “HPE transformer: Learning to optimize multi-group multicast beamforming under nonconvex QoS constraints,”IEEE Trans. Commun., vol. 72, no. 9, pp. 5581–5594, Sep. 2024 on Pith
- [33] J. Zhang, M. Liu, J. Tang, N. Zhao, D. Niyato, and X. Wang, “Joint design for RIS-aided ISAC via deep unfolding learning,”IEEE Trans. Cogn. Commun. Netw., vol. 11, no. 1, pp. 349–361, Feb. 2025 on Pith
- [34] Q. Gao, R. Zhong, Y . Zou, H. Shin, and Y . Liu, “Deep learning- based beamforming optimization for isac systems: A low-complexity and transferable framework,”IEEE Trans. Wireless Commun., vol. 25, pp. 8754–8768, 2026 on Pith
- [35] X. Zhang, W. Yuan, C. Liu, J. Wu, and D. W. K. Ng, “Predictive beamforming for vehicles with complex behaviors in ISAC systems: A deep learning approach,”IEEE J. Sel. Topics Signal Process., vol. 18, no. 5, pp. 828–841, Jul. 2024 on Pith
- [36] X. Yu, J. Xu, C. Dong, C. Xing, N. Zhao, Q. Wu, and D. Niyato, “Learning-based predictive beamforming for secure ISAC via IRS,” IEEE Trans. Commun., vol. 73, no. 12, pp. 14 304–14 315, Dec. 2025 on Pith
- [37] M. Zhang, R. Zhong, X. Mu, H. Shin, and Y . Liu, “AI-enabled near-field communications: User movement prediction and beam tracking,”IEEE Trans. Cogn. Commun. Netw., vol. 12, pp. 6247–6264, 2026 on Pith
- [38] P. Jiang, M. Li, R. Liu, and Q. Liu, “Graph learning for cooperative cell-free ISAC systems: From optimization to estimation,”IEEE Trans. Wireless Commun., vol. 25, pp. 13 992–14 008, 2026 on Pith
- [39] M. Cui and L. Dai, “Channel estimation for extremely large-scale MIMO: Far-field or near-field?”IEEE Trans. Commun., vol. 70, no. 4, pp. 2663–2677, Apr. 2022 on Pith
- [40] C. Wu, C. You, Y . Liu, L. Chen, and S. Shi, “Two-stage hierarchical beam training for near-field communications,”IEEE Trans. Veh. Tech- nol., vol. 73, no. 2, pp. 2032–2044, Feb. 2024 on Pith
- [41] H. Hua, J. Xu, and Y . C. Eldar, “Near-field 3D localization via MIMO radar: Cram ´er–rao bound analysis and estimator design,”IEEE Trans. Signal Process., vol. 72, pp. 3879–3895, 2024 on Pith
- [42] A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,”Advances in Neural Information Processing Systems (NeurIPS), vol. 30, 2017 on Pith
- [43] C. B. Peel, B. M. Hochwald, and A. L. Swindlehurst, “A vec- tor-perturbation technique for near-capacity multiantenna multiuser communication—part i: Channel inversion and regularization,”IEEE Trans. Commun., vol. 53, no. 1, pp. 195–202, Jan. 2005 on Pith
- [44] J. Nocedal and S. J. Wright,Numerical Optimization, 2nd ed. New York, NY , USA: Springer, 2006 on Pith
Challenges and discussion
Disagree with the review, or know something it missed? Leave a comment, or challenge this review formally: a challenge becomes a permanent entry in this paper's record, and its decision is published here.
Come back by email
A short list of the week's newly reviewed papers. Nothing is sent until you confirm from your inbox, and every email has a link that stops it. No account.
No comments yet.
Sign in to comment