pith. sign in

arxiv: 2604.26414 · v1 · submitted 2026-04-29 · 📡 eess.SP

A Novel Reinforcement Learning Based Framework for Scalable MIMO Interference Alignment

Samitha Gunarathne , Eslam Eldeeb , Nurul Huda Mahmood , Italo Atzeni This is my paper

Pith reviewed 2026-05-07 10:49 UTC · model grok-4.3

classification 📡 eess.SP
keywords interference alignmentreinforcement learningMIMO networksthroughput optimizationCSI estimationsubspace coordinationmulti-user MIMOscalability
0
0 comments X

The pith

A reinforcement learning framework achieves scalable interference alignment in large MIMO systems without global CSI by learning subspace coordination.

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

The paper develops an IA-inspired learning algorithm to maximize network throughput in multi-user MIMO setups while addressing the barriers of global CSI requirements and intractable closed-form solutions. For small-scale systems it introduces a predictive transformer model that estimates CSI and lowers signaling overhead. For large-scale systems the IA task is recast as a multi-objective optimization over subspace coordination, then solved by two new reinforcement learning algorithms that train agents to align interference at the receivers. Simulation results show these methods produce up to 30 percent higher average user throughput than the strongest conventional baselines.

Core claim

By training reinforcement learning agents to select transmit and receive subspaces in a distributed fashion, the framework performs interference alignment at scale, removes the need for global channel state information at each transmitter, and yields substantial throughput gains in large MIMO networks where analytical solutions are unavailable.

What carries the argument

Reinforcement learning agents that learn policies for coordinated subspace selection to align interference without global CSI.

If this is right

  • Throughput rises by up to 30 percent over conventional baselines in simulated large-scale MIMO deployments.
  • Signaling overhead drops in small MIMO systems through predictive CSI estimation instead of frequent full feedback.
  • The method extends interference alignment to network sizes where closed-form solutions become intractable.
  • A data-driven alternative replaces the need to derive analytical IA precoders for complex MIMO configurations.

Where Pith is reading between the lines

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

  • If the learned policies generalize, the same RL structure could be reused for other distributed interference-management tasks in wireless networks.
  • Combining the transformer CSI estimator with additional sensor data might further cut feedback requirements in high-mobility scenarios.
  • The multi-objective formulation opens the door to adding explicit fairness or energy constraints into the same learning loop.

Load-bearing premise

Reinforcement learning agents trained only in simulation will transfer successfully to real MIMO channels and the transformer CSI estimator will stay accurate enough across changing mobility and interference conditions.

What would settle it

A hardware testbed experiment or field trial that measures end-to-end user throughput under realistic time-varying channels and compares the achieved gains against the 30 percent simulation figure.

Figures

Figures reproduced from arXiv: 2604.26414 by Eslam Eldeeb, Italo Atzeni, Nurul Huda Mahmood, Samitha Gunarathne.

Figure 1
Figure 1. Figure 1: Multi-user MIMO system model with K transmitters and receivers. modeled as a time-correlated static Rayleigh fading MIMO channel. All desired transmissions occur between pairs with the same index, i.e., the k-th transmitter communicates with the k-th receiver. Consequently, all other received signals are considered as interference. The received signal at the j-th user and t-th transmission interval (yj (t)… view at source ↗
Figure 2
Figure 2. Figure 2: Timeline of the TF-IA scheme. data, i.e., Hj,i, are reshaped to serve as input to the transformer model as             Re[h (1) j,i (t)] Re[h (1) j,i (t − 1)] · · · Re[h (1) j,i (t − T )] . . . . . . . . . . . . Re[h (mt) j,i (t)] Re[h (mt) j,i (t − 1)] · · · Re[h (mt) j,i (t − T )] Im[h (1) j,i (t)] Im[h (1) j,i (t − 1)] · · · Im[h (1) j,i (t − T )] . . . . . . . . . . . . Im[h (mt) j,i (t)] I… view at source ↗
Figure 3
Figure 3. Figure 3: Subspace decomposition at the receiver under the pro view at source ↗
Figure 4
Figure 4. Figure 4: Variation of sum rate with SNR for a MIMO system with view at source ↗
Figure 5
Figure 5. Figure 5: Variation of sum rate with SNR for a MIMO system with view at source ↗
Figure 6
Figure 6. Figure 6: Convergence of the proposed RL-MaxSNR approach for a view at source ↗
Figure 7
Figure 7. Figure 7: Distance variation of the proposed RL-MaxSNR model a view at source ↗
Figure 8
Figure 8. Figure 8: Convergence of the throughput of proposed RL-MaxSNR view at source ↗
Figure 9
Figure 9. Figure 9: Variation of average user throughput with SNR for a MI view at source ↗
Figure 11
Figure 11. Figure 11: Impact of the proposed schemes on average user throu view at source ↗
read the original abstract

Interference alignment (IA) is a widely recognized approach for mitigating inter-cell interference in multi-user multiple-input multiple-output (MIMO) networks. Despite its effectiveness, practical deployment remains constrained by two major challenges, i.e., the need for global channel state information (CSI) at each transmitter and the complexity of deriving closed-form solutions for intricate MIMO systems. This work aims to maximize network throughput by effectively mitigating interference using an IA-inspired learning algorithm that addresses its aforementioned challenges. First, we propose a predictive, transformer-based IA framework that estimates CSI to reduce signaling overhead in small-scale MIMO systems. Next, we formulate the IA problem as a multi-objective optimization problem based on subspace coordination and develop two reinforcement learning-based algorithms to enhance the scalability of IA in large-scale MIMO systems. Simulation results demonstrate that the proposed methods significantly outperform conventional baselines with up to 30% average user throughput gains over the best performing baseline.

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

3 major / 2 minor

Summary. The manuscript proposes a predictive transformer-based framework for CSI estimation to reduce signaling overhead in small-scale MIMO interference alignment, and formulates IA as a multi-objective subspace-coordination problem solved by two reinforcement learning algorithms for scalability in large-scale MIMO. Simulation results are presented claiming that the methods significantly outperform conventional baselines, with up to 30% average user throughput gains.

Significance. If the reported gains prove robust and the learned policies generalize, the work could offer a practical path to scalable IA without requiring global CSI or closed-form solutions, addressing longstanding deployment barriers in multi-cell MIMO networks. The combination of transformer-based prediction with RL-driven subspace alignment represents a coherent learning-based alternative to traditional IA, though its significance remains provisional given the simulation-only evidence.

major comments (3)
  1. [Simulation Results] The headline performance claim of up to 30% average user throughput gains (Abstract and Simulation Results) rests on simulations whose setup details, baseline implementations, statistical significance testing, error bars, and ablation studies are not visible, preventing independent verification of whether the gains are robust or sensitive to hyperparameter choices.
  2. [Reinforcement Learning Algorithms] The RL algorithms are trained and evaluated inside the identical simulated environment used to define the optimization objectives (Reinforcement Learning Algorithms section), creating a circularity risk; no cross-validation on alternate channel distributions (e.g., spatially correlated, Rician, or time-varying Doppler models) is reported, so the subspace-coordination policies may overfit the training distribution rather than generalize.
  3. [Transformer-based IA Framework] The transformer-based CSI estimator is presented without Doppler-sensitivity analysis or tests under mobility and dynamic interference that violate the training distribution (Transformer-based IA Framework section), leaving open whether prediction accuracy remains sufficient for the claimed overhead reduction in realistic conditions.
minor comments (2)
  1. [Problem Formulation] Notation for the multi-objective optimization (e.g., weighting between sum-rate and interference leakage terms) should be clarified with an explicit equation reference to avoid ambiguity when comparing the two RL variants.
  2. [Simulation Results] Figure captions for throughput CDFs and convergence plots would benefit from explicit mention of the number of Monte Carlo runs and channel realizations used.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which have helped strengthen the presentation of our work. We address each major comment point by point below, indicating where revisions have been made to the manuscript.

read point-by-point responses

  1. Referee: [Simulation Results] The headline performance claim of up to 30% average user throughput gains (Abstract and Simulation Results) rests on simulations whose setup details, baseline implementations, statistical significance testing, error bars, and ablation studies are not visible, preventing independent verification of whether the gains are robust or sensitive to hyperparameter choices.

    Authors: We agree that the original manuscript did not provide sufficient detail on the simulation setup to enable independent verification. In the revised version, the Simulation Results section has been substantially expanded to include: explicit descriptions of all baseline implementations (including code-level parameters where applicable), error bars computed across 100 independent Monte Carlo runs with different random seeds, paired t-test results confirming statistical significance of the throughput gains, and ablation studies varying key hyperparameters such as transformer depth, RL discount factor, and number of training episodes. These additions confirm that the reported gains remain consistent within the evaluated parameter ranges. revision: yes

  2. Referee: [Reinforcement Learning Algorithms] The RL algorithms are trained and evaluated inside the identical simulated environment used to define the optimization objectives (Reinforcement Learning Algorithms section), creating a circularity risk; no cross-validation on alternate channel distributions (e.g., spatially correlated, Rician, or time-varying Doppler models) is reported, so the subspace-coordination policies may overfit the training distribution rather than generalize.

    Authors: The referee correctly highlights a limitation in generalization testing. While the core training and evaluation occur within the same i.i.d. Rayleigh fading model used to formulate the objectives, we have added cross-validation experiments in the revised manuscript using spatially correlated channels (exponential correlation model) and Rician fading with varying K-factors. The subspace-coordination policies retain positive throughput gains under these distributions, although the magnitude is reduced compared to the original setting. We acknowledge that time-varying Doppler models were not tested and have added an explicit discussion of this as a limitation with suggested directions for future extension. revision: partial

  3. Referee: [Transformer-based IA Framework] The transformer-based CSI estimator is presented without Doppler-sensitivity analysis or tests under mobility and dynamic interference that violate the training distribution (Transformer-based IA Framework section), leaving open whether prediction accuracy remains sufficient for the claimed overhead reduction in realistic conditions.

    Authors: We agree that the absence of mobility and Doppler analysis limits claims about practical overhead reduction. The original framework targeted quasi-static small-scale MIMO scenarios. In the revision, we have added a dedicated Doppler-sensitivity subsection and corresponding simulations under time-varying channels with user mobility (Jakes' model at different maximum Doppler frequencies). The results show that prediction accuracy and resulting throughput gains hold for low-to-moderate Doppler spreads but degrade at high mobility; the overhead reduction benefit persists up to a quantifiable threshold. These analyses are now included in the Transformer-based IA Framework section. revision: yes

Circularity Check

0 steps flagged

No circularity: simulation-based RL evaluation is independent of derivation inputs

full rationale

The paper defines a transformer CSI estimator and formulates IA as a multi-objective subspace coordination problem, then applies RL agents to solve it. Reported throughput gains are obtained by running the trained agents in the same simulator used for training. This is standard empirical validation rather than a reduction by construction: the objective function and channel model are fixed inputs, the RL policy is learned, and the numerical gains are an output of executing that policy. No equations or text in the provided sections equate the final performance metric to the training objective by definition, no self-citations are load-bearing for the central claim, and no ansatz or uniqueness theorem is smuggled in. The derivation chain remains self-contained against external benchmarks.

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 RL reward functions and transformer training objectives are implicitly assumed to align with throughput maximization but details are absent.

pith-pipeline@v0.9.0 · 5465 in / 1097 out tokens · 44430 ms · 2026-05-07T10:49:13.689077+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

52 extracted references · 2 canonical work pages

  1. [1]

    Framework and overall objectives of the future d evelopment of international mobile telecommunications (IMT) for 2030 an d beyond,

    ITU-R, “Framework and overall objectives of the future d evelopment of international mobile telecommunications (IMT) for 2030 an d beyond,” International Telecommunication Union (ITU) Recommendat ion (ITU- R), 2023

    2030

  2. [2]

    The roadmap to 6G: AI empowered wireless networks,

    K. B. Letaief et al. , “The roadmap to 6G: AI empowered wireless networks,” IEEE Commun. Mag. , vol. 57, no. 8, pp. 84–90, 2019

    2019

  3. [3]

    AI-empowered multiple access for 6G: A survey of spectrum sensing, protocol designs, and optimizations,

    X. Cao et al. , “AI-empowered multiple access for 6G: A survey of spectrum sensing, protocol designs, and optimizations,” Proceedings of the IEEE , vol. 112, no. 9, pp. 1264–1302, 2024

    2024

  4. [4]

    White paper on broadband connectivity in 6G,

    N. Rajatheva et al. , “White paper on broadband connectivity in 6G,”

  5. [5]

    Available: https://arxiv.org/abs/2004

    [Online]. Available: https://arxiv.org/abs/2004. 14247

    2004

  6. [6]

    Machine type communications: key drivers and enablers towards the 6G era,

    N. H. Mahmood et al. , “Machine type communications: key drivers and enablers towards the 6G era,” EURASIP J. Wireless Comm. and Netw. , no. 134, 2021. [Online]. Available: https: //doi.org/10.1186/s13638-021-02010-5

    work page doi:10.1186/s13638-021-02010-5 2021

  7. [7]

    6G white paper on machine learning in wireless communication networks,

    S. Ali et al. , “6G white paper on machine learning in wireless communication networks,” 2020. [Online]. Available: http s://arxiv.org/ abs/2004.13875

    work page arXiv 2020

  8. [8]

    A survey on machine learning techniques fo r massive MIMO configurations: Application areas, performance limit ations and future challenges,

    P . K. Gkonis, “A survey on machine learning techniques fo r massive MIMO configurations: Application areas, performance limit ations and future challenges,” IEEE Access , vol. 11, pp. 67–88, 2023

    2023

  9. [9]

    6G wireless systems: A vision, architectural elements, and future directions,

    L. U. Khan et al., “6G wireless systems: A vision, architectural elements, and future directions,” IEEE Access, vol. 8, pp. 147 029–147 044, 2020

    2020

  10. [10]

    Future of ultra-dense networks beyond 5G: Harnessing heterogeneous moving cells,

    S. Andreev, V . Petrov, M. Dohler, and H. Y anikomeroglu, “ Future of ultra-dense networks beyond 5G: Harnessing heterogeneous moving cells,” IEEE Commun. Mag. , vol. 57, no. 6, pp. 86–92, 2019

    2019

  11. [11]

    Massive MIMO for next generation wireless systems,

    E. G. Larsson, O. Edfors, F. Tufvesson, and T. L. Marzett a, “Massive MIMO for next generation wireless systems,” IEEE Commun. Mag. , vol. 52, no. 2, pp. 186–195, 2014

    2014

  12. [12]

    Interference alignment and its applications: A survey, research issues, and challenges,

    N. Zhao et al. , “Interference alignment and its applications: A survey, research issues, and challenges,” IEEE Commun. Surveys Tuts. , vol. 18, no. 3, pp. 1779–1803, 2016. 14

    2016

  13. [13]

    A distribute d numerical approach to interference alignment and applications to wir eless interfer- ence networks,

    K. Gomadam, V . R. Cadambe, and S. A. Jafar, “A distribute d numerical approach to interference alignment and applications to wir eless interfer- ence networks,” IEEE Trans. Inf. Theory , vol. 57, no. 6, pp. 3309–3322, 2011

    2011

  14. [14]

    Feedback-topology designs for interference alignment in MIMO interference channels,

    S. Cho et al. , “Feedback-topology designs for interference alignment in MIMO interference channels,” IEEE Trans. Signal Process. , vol. 60, no. 12, pp. 6561–6575, 2012

    2012

  15. [15]

    A rapid convergent low complexity interference align- ment algorithm for wireless sensor networks,

    L. Jiang et al. , “A rapid convergent low complexity interference align- ment algorithm for wireless sensor networks,” Sensors, vol. 15, no. 8, pp. 18 526–18 549, 2015

    2015

  16. [16]

    Linear trans ceiver design for interference alignment: Complexity and computation,

    M. Razaviyayn, M. Sanjabi, and Z.-Q. Luo, “Linear trans ceiver design for interference alignment: Complexity and computation,” IEEE Trans. Inf. Theory , vol. 58, no. 5, pp. 2896–2910, 2012

    2012

  17. [17]

    Interference a lignment as a rank constrained rank minimization,

    D. S. Papailiopoulos and A. G. Dimakis, “Interference a lignment as a rank constrained rank minimization,” IEEE Trans. Signal Process. , vol. 60, no. 8, pp. 4278–4288, 2012

    2012

  18. [18]

    Recurrent neural networks: vanishing and exploding gradients are not the end of the story,

    N. Zucchet and A. Orvieto, “Recurrent neural networks: vanishing and exploding gradients are not the end of the story,” Advances in Neural Information Processing Systems , vol. 37, pp. 139 402–139 443, 2024

    2024

  19. [19]

    Transf ormer- aided CSI prediction for interference alignment in MIMO sys tems,

    S. Gunarathne, N. H. Mahmood, and M. Latva-aho, “Transf ormer- aided CSI prediction for interference alignment in MIMO sys tems,” in EuCNC/6G Summit , 2025, pp. 13–18

    2025

  20. [20]

    Cooperative algorithms fo r MIMO interference channels,

    S. W. Peters and R. W. Heath, “Cooperative algorithms fo r MIMO interference channels,” IEEE Trans. V eh. Technol. , vol. 60, no. 1, pp. 206–218, 2011

    2011

  21. [21]

    An iterative interference alignm ent algorithm for the general MIMO X channel,

    Y . Wei and T.-M. Lok, “An iterative interference alignm ent algorithm for the general MIMO X channel,” IEEE Trans. Wireless Commun., vol. 18, no. 3, pp. 1847–1859, 2019

    2019

  22. [22]

    Co mmuni- cation over MIMO X channels: Interference alignment, decom position, and performance analysis,

    M. A. Maddah-Ali, A. S. Motahari, and A. K. Khandani, “Co mmuni- cation over MIMO X channels: Interference alignment, decom position, and performance analysis,” IEEE Trans. Inf. Theory , vol. 54, no. 8, pp. 3457–3470, 2008

    2008

  23. [24]

    Interference alignm ent for MIMO downlink multicell networks,

    W. Liu, J.-X. Sun, J. Li, and Y . Ma, “Interference alignm ent for MIMO downlink multicell networks,” IEEE Trans. V eh. Technol., vol. 65, no. 8, pp. 6159–6167, 2016

    2016

  24. [25]

    Interference alignment te chniques for MIMO multi-cell interfering broadcast channels,

    J. Tang and S. Lambotharan, “Interference alignment te chniques for MIMO multi-cell interfering broadcast channels,” IEEE Trans. Com- mun., vol. 61, no. 1, pp. 164–175, 2013

    2013

  25. [26]

    Performance analysis and optimiza tion for interference alignment over MIMO interference channels wi th limited feedback,

    X. Chen and C. Y uen, “Performance analysis and optimiza tion for interference alignment over MIMO interference channels wi th limited feedback,” IEEE Trans. Signal Process. , vol. 62, no. 7, pp. 1785–1795, 2014

    2014

  26. [27]

    Enhanced AI-based CSI prediction solutions for massive MIMO in 5G and 6G systems,

    D. Burghal et al. , “Enhanced AI-based CSI prediction solutions for massive MIMO in 5G and 6G systems,” IEEE Access , vol. 11, pp. 117 810–117 825, 2023

    2023

  27. [28]

    Machine learning- based channel prediction in massive MIMO with channel aging,

    J. Y uan, H. Q. Ngo, and M. Matthaiou, “Machine learning- based channel prediction in massive MIMO with channel aging,” IEEE Trans. Wireless Commun., vol. 19, no. 5, pp. 2960–2973, 2020

    2020

  28. [29]

    Deep learning for fading ch annel prediction,

    W. Jiang and H. D. Schotten, “Deep learning for fading ch annel prediction,” IEEE Open J. Commun. Soc. , vol. 1, pp. 320–332, 2020

    2020

  29. [30]

    Transformer network based channel prediction for CSI feedback enhancement in AI-native air interface,

    T. Zhou et al. , “Transformer network based channel prediction for CSI feedback enhancement in AI-native air interface,” IEEE Trans. Wireless Commun., vol. 23, no. 9, pp. 11 154–11 167, 2024

    2024

  30. [31]

    Channel state information prediction for 5G wireless communications: A deep learning approach,

    C. Luo et al. , “Channel state information prediction for 5G wireless communications: A deep learning approach,” IEEE Trans. Netw. Sci. Eng., vol. 7, no. 1, pp. 227–236, 2020

    2020

  31. [32]

    Transform er- empowered predictive beamforming for rate-splitting mult iple access in non-terrestrial networks,

    S. Zhang, S. Zhang, W. Y uan, and T. Q. S. Quek, “Transform er- empowered predictive beamforming for rate-splitting mult iple access in non-terrestrial networks,” IEEE Trans. Wireless Commun. , vol. 23, no. 12, pp. 19 776–19 788, 2024

    2024

  32. [33]

    Transformer-based channel prediction for rate-splitti ng multiple access-enabled vehicle-to-everything communic ation,

    S. Zhang et al., “Transformer-based channel prediction for rate-splitti ng multiple access-enabled vehicle-to-everything communic ation,” IEEE Trans. Wireless Commun. , vol. 23, no. 10, pp. 12 717–12 730, 2024

    2024

  33. [34]

    An offline multi-agent reinforc ement learn- ing framework for radio resource management,

    E. Eldeeb and H. Alves, “An offline multi-agent reinforc ement learn- ing framework for radio resource management,” IEEE Trans. Mobile Comput., pp. 1–14, 2025

    2025

  34. [35]

    Multi-UA V path learning for age and power opti- mization in IoT with UA V battery recharge,

    E. Eldeeb et al. , “Multi-UA V path learning for age and power opti- mization in IoT with UA V battery recharge,” IEEE Trans. V eh. Technol., vol. 72, no. 4, pp. 5356–5360, 2022

    2022

  35. [36]

    Spectrum sharing in vehicu lar networks based on multi-agent reinforcement learning,

    L. Liang, H. Y e, and G. Y . Li, “Spectrum sharing in vehicu lar networks based on multi-agent reinforcement learning,” IEEE J. Sel. Areas Com- mun., vol. 37, no. 10, pp. 2282–2292, 2019

    2019

  36. [37]

    Joint interference alignment and power control for dense networks via deep reinforcement learning,

    C. Wang et al., “Joint interference alignment and power control for dense networks via deep reinforcement learning,” IEEE Wireless Commun. Lett., vol. 10, no. 5, pp. 966–970, 2021

    2021

  37. [38]

    Reinforcement learning with selective exploration for interference management in mmWave networks,

    S. Dinh-van et al. , “Reinforcement learning with selective exploration for interference management in mmWave networks,” IEEE Trans. Mach. Learn. Commun. Netw. , vol. 3, pp. 280–295, 2025

    2025

  38. [39]

    Reinforcement learning-based downlink interference control for ultra-dense small cells,

    L. Xiao et al. , “Reinforcement learning-based downlink interference control for ultra-dense small cells,” IEEE Trans. Wireless Commun. , vol. 19, no. 1, pp. 423–434, 2020

    2020

  39. [40]

    On feas ibility of interference alignment in MIMO interference networks,

    C. M. Y etis, T. Gou, S. A. Jafar, and A. H. Kayran, “On feas ibility of interference alignment in MIMO interference networks,” IEEE Trans. Signal Process. , vol. 58, no. 9, pp. 4771–4782, 2010

    2010

  40. [41]

    M. A. Maddah-Ali, A. K. Khandani, and M. A. Sadrabadi, An efficient signaling scheme for MIMO broadcast systems: Design and per formance evaluation. Department of Electrical and Computer Engineering, University of Waterloo, 2005

    2005

  41. [42]

    Inter- cell interference sub-space coordination for 5G ultra-den se networks,

    A. Karimi, N. H. Mahmood, K. I. Pedersen, and P . Mogensen , “Inter- cell interference sub-space coordination for 5G ultra-den se networks,” in Proc. IEEE V eh. Technol. Conf. , 2017, pp. 1–5

    2017

  42. [43]

    A leakage-based MMSE beamfor ming design for a MIMO interference channel,

    F. Sun and E. de Carvalho, “A leakage-based MMSE beamfor ming design for a MIMO interference channel,” IEEE Signal Process. Lett. , vol. 19, no. 6, pp. 368–371, 2012

    2012

  43. [44]

    A novel upli nk MIMO transmission scheme in a multicell environment,

    B. O. Lee, H. W. Je, O.-S. Shin, and K. B. Lee, “A novel upli nk MIMO transmission scheme in a multicell environment,” IEEE Trans. Wireless Commun., vol. 8, no. 10, pp. 4981–4987, 2009

    2009

  44. [45]

    Interference alignment vi a alternating minimization,

    S. W. Peters and R. W. Heath, “Interference alignment vi a alternating minimization,” in Proc. IEEE Int. Conf. Acoust., Speech, Signal Pro- cessing, 2009, pp. 2445–2448

    2009

  45. [46]

    Interference al ignment schemes for K-user interference channel based on manifold o ptimiza- tion,

    C. Zhang, Z. Liu, T. Hong, and G. Zhang, “Interference al ignment schemes for K-user interference channel based on manifold o ptimiza- tion,” EURASIP J. Wireless Comm. and Netw. , vol. 2019, no. 1, p. 196, Aug 2019

    2019

  46. [47]

    An iterativ ely weighted MMSE approach to distributed sum-utility maximization for a MIMO interfering broadcast channel,

    Q. Shi, M. Razaviyayn, Z.-Q. Luo, and C. He, “An iterativ ely weighted MMSE approach to distributed sum-utility maximization for a MIMO interfering broadcast channel,” IEEE Trans. Signal Process. , vol. 59, no. 9, pp. 4331–4340, 2011

    2011

  47. [48]

    The practical challenges of interference alignment,

    O. El Ayach, S. W. Peters, and R. W. Heath, “The practical challenges of interference alignment,” IEEE Wireless Commun. , vol. 20, no. 1, pp. 35–42, 2013

    2013

  48. [49]

    Attention is all you need,

    A. V aswani et al. , “Attention is all you need,” Advances in neural information processing systems , vol. 30, 2017

    2017

  49. [50]

    A survey of transform ers,

    T. Lin, Y . Wang, X. Liu, and X. Qiu, “A survey of transform ers,” AI Open, vol. 3, pp. 111–132, 2022

    2022

  50. [51]

    R. S. Sutton, A. G. Barto et al., Reinforcement learning: An introduction. MIT press Cambridge, 1998, vol. 1, no. 1

    1998

  51. [52]

    Continuous control with deep reinforcement learning,

    T. P . Lillicrap et al. , “Continuous control with deep reinforcement learning,” arXiv: Learning , 2015. [Online]. Available: https://api. semanticscholar.org/CorpusID:16326763

    2015

  52. [53]

    Novel learning-based multiuser detection algo- rithms for spatially correlated MTC,

    T. Sivalingam et al. , “Novel learning-based multiuser detection algo- rithms for spatially correlated MTC,” IEEE Internet of Things J. , vol. 12, no. 13, pp. 23 169–23 181, 2025

    2025