arxiv: 2604.04589 · v1 · submitted 2026-04-06 · 💻 cs.AI · cs.LG

Recognition: 2 theorem links

· Lean Theorem

Greedy and Transformer-Based Multi-Port Selection for Slow Fluid Antenna Multiple Access

Darian Perez-Adan , Jose P. Gonzalez-Coma , F. Javier Lopez-Martinez , Luis Castedo

Authors on Pith no claims yet

Pith reviewed 2026-05-10 19:25 UTC · model grok-4.3

classification 💻 cs.AI cs.LG

keywords fluid antennaport selectionmultiple accessspectral efficiencygreedy algorithmtransformer networkimitation learning

0 comments

The pith

A greedy forward-selection method with swaps and a Transformer network solve port selection in fluid antenna multiple access for better spectral efficiency or lower cost.

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

The paper focuses on choosing which ports to activate on multi-port fluid antennas in slow fluid antenna multiple access systems to maximize spectral efficiency. It introduces GFwd+S, a greedy forward-selection procedure followed by swap refinement, that beats existing reference methods in spectral efficiency, along with a Transformer neural network trained first by imitation learning and then refined with a policy-gradient stage that reaches nearly the same performance but runs faster. A reader would care because fluid antennas adapt their physical shape or port usage to the radio channel, offering a way to improve multi-user wireless links without extra hardware. If these selection strategies hold up, they make such adaptive systems more practical by resolving the usual conflict between high performance and feasible computation. The work stays within simulated channel models for its comparisons.

Core claim

Existing port-selection methods for fluid antenna multiple access either reach near-optimal spectral efficiency at high computational cost or lose substantial performance for the sake of speed. GFwd+S, a greedy forward-selection algorithm with a swap-refinement step, consistently outperforms the reference schemes in spectral efficiency. A Transformer-based neural network trained via imitation learning followed by a Reinforce policy-gradient stage approaches the same spectral efficiency level while using far less computation.

What carries the argument

GFwd+S greedy forward-selection with swap refinement, together with the Transformer neural network for port selection in multi-port fluid antennas.

If this is right

GFwd+S raises spectral efficiency above that of prior port-selection schemes in the tested multi-user FAMA settings.
The Transformer network delivers comparable spectral efficiency at substantially reduced computational cost.
Imitation learning followed by policy-gradient refinement produces a neural network that learns effective port-selection policies from the greedy method.
The two approaches together address the performance-complexity trade-off that limited earlier solutions in slow FAMA.

Where Pith is reading between the lines

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

Hardware experiments with real fluid antennas would test whether the simulated gains survive hardware impairments and imperfect channel knowledge.
The selection logic might carry over to other reconfigurable antenna systems that must pick subsets of elements or beams under similar constraints.
Lower-complexity neural solutions could allow port selection to run on resource-limited devices in dynamic wireless environments.

Load-bearing premise

The simulation channel models and multi-user scenarios used to measure spectral efficiency are representative enough that the reported gains will translate to real deployments.

What would settle it

A measurement campaign with physical fluid-antenna hardware in a multi-user testbed that checks whether the simulated spectral-efficiency gains over the reference schemes actually appear.

Figures

Figures reproduced from arXiv: 2604.04589 by Darian Perez-Adan, F. Javier Lopez-Martinez, Jose P. Gonzalez-Coma, Luis Castedo.

**Figure 3.** Figure 3: presents the average SE versus the transmit SNR. Standalone GFwd slightly improves upon GEPort while requiring lower complexity. With swap refinement, GFwd+S consistently outperforms GEPort by up to roughly 40%, confirming that swap refinement is essential for incremental forward construction to overcome the suboptimal early decisions inherent to backward elimination. Recall that Proposition 1 in Append… view at source ↗

**Figure 4.** Figure 4: shows the average SE versus the number of swap rounds R for SNR values in {10, 15, 20} dB. A single swap round recovers most of the gain over GFwd, with only marginal improvement beyond R = 2. In addition, GFwd+S consistently outperforms GEPort for all considered SNR values and R ≥ 1, confirming that a small number of swap rounds (e.g., R = 3) is enough to converge at a stable solution. 0/GFwd 1 2 3 4 5 3… view at source ↗

**Figure 6.** Figure 6: depicts the SE versus the number of active ports L. All methods benefit from increasing L due to additional combining gain. GFwd+S leads to the highest SE, while the proposed NN outperforms GEPort for L ≥ 8. The gap between GFwd+S and GEPort is largest for intermediate values of L (6–12), where port selection is most combinatorial, and narrows for very small or very large L. 2 4 6 8 10 12 14 16 0 2 4 6 Num… view at source ↗

**Figure 7.** Figure 7: shows the SE versus the total number of ports P at SNR = 15 dB. Increasing P with fixed aperture W = 4λ densifies the port grid and increases the spatial correlation among ports. In this regime, CUMA, which does not jointly account for Ak and Bk degrades, while GFwd+S and the proposed NN maintain their advantage, as observed in [6]. The NN also consistently outperforms GEPort across all P values, achieving… view at source ↗

read the original abstract

We address the port-selection problem in fluid antenna multiple access (FAMA) systems with multi-port fluid antenna (FA) receivers. Existing methods either achieve near-optimal spectral efficiency (SE) at prohibitive computational cost or sacrifice significant performance for lower complexity. We propose two complementary strategies: (i) GFwd+S, a greedy forward-selection method with swap refinement that consistently outperforms state-of-the-art reference schemes in terms of SE, and (ii) a Transformer-based neural network trained via imitation learning followed by a Reinforce policy-gradient stage, which approaches GFwd+S performance at lower computational cost.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper gives two concrete algorithms for multi-port selection in slow fluid-antenna multiple access, but the abstract supplies no numbers or setup details to back the performance claims.

read the letter

The core contribution is a greedy forward-selection routine with a swap refinement step (GFwd+S) plus a transformer network trained first by imitation learning and then by a Reinforce stage. Both target the problem of picking ports from a fluid antenna to maximize spectral efficiency in a multi-user setting while keeping complexity manageable. The greedy-plus-swap idea and the specific two-stage training for the transformer look new relative to the methods the abstract cites. The work frames the trade-off between near-optimal but expensive selection and cheaper but weaker alternatives in a practical way, which is useful for people thinking about transceiver design in fluid-antenna systems. The abstract states that GFwd+S beats existing references and that the transformer comes close at lower cost, yet it gives no quantitative results, no channel model description, no error bars, and no simulation parameters. That makes it impossible to judge whether the reported gains are real, how large they are, or whether they survive changes in user geometry or correlation structure. The stress-test concern about unverified fidelity of the simulated channels to real deployments is fair; nothing in the abstract shows sensitivity checks or alternative models. If the full manuscript contains solid Monte-Carlo results with realistic parameters and reproducible code, the methods could be worth following. Right now the evidence is too thin to assess impact. This is for readers already working on fluid-antenna or reconfigurable-antenna problems who want algorithmic options. A serious referee should see the full simulations and derivations before deciding on publication.

Referee Report

1 major / 1 minor

Summary. The paper addresses the multi-port selection problem in slow fluid antenna multiple access (FAMA) systems. It proposes GFwd+S, a greedy forward-selection algorithm with swap refinement claimed to consistently outperform state-of-the-art reference schemes in spectral efficiency (SE), and a Transformer-based neural network trained via imitation learning followed by a Reinforce policy-gradient stage that approaches GFwd+S performance at lower computational cost.

Significance. If the reported SE gains hold under representative channel conditions, the work could advance practical FAMA deployments by offering a high-performance low-complexity alternative to existing port-selection methods, particularly for multi-user scenarios where computational efficiency matters.

major comments (1)

[Numerical Results / Simulation Setup] The central outperformance claim for GFwd+S and the Transformer relies on Monte-Carlo SE evaluations under a fixed multi-user channel model. No sensitivity study, alternative geometry-based model, or hardware validation is indicated, leaving the mapping from simulation statistics (spatial correlation, slow fluid motion, interference) to real deployments untested and load-bearing for the broader assertion of consistent gains.

minor comments (1)

[Abstract] The abstract states performance claims (outperformance, lower cost) without any quantitative SE values, error bars, or simulation parameters, reducing clarity for readers.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address the major comment on the numerical results and simulation setup below, and we outline planned revisions to strengthen the work.

read point-by-point responses

Referee: The central outperformance claim for GFwd+S and the Transformer relies on Monte-Carlo SE evaluations under a fixed multi-user channel model. No sensitivity study, alternative geometry-based model, or hardware validation is indicated, leaving the mapping from simulation statistics (spatial correlation, slow fluid motion, interference) to real deployments untested and load-bearing for the broader assertion of consistent gains.

Authors: We appreciate the referee's emphasis on robustness. The channel model used follows the standard formulation in the FAMA literature, explicitly incorporating spatial correlation across ports, slow fluid motion dynamics, and multi-user interference. Monte-Carlo averaging over a large number of independent realizations provides statistical reliability for the reported SE gains. We agree that sensitivity analysis would further support the claims. In the revised manuscript we will add new simulation results varying the number of users, spatial correlation strength, fluid antenna size, and interference levels, together with additional discussion mapping these parameters to practical deployment scenarios. Hardware validation lies outside the scope of this theoretical and algorithmic study. revision: partial

Circularity Check

0 steps flagged

No circularity: algorithmic proposals and simulation-based comparisons are self-contained

full rationale

The paper introduces GFwd+S as a greedy forward-selection algorithm with swap refinement and a Transformer network trained first by imitation learning then by Reinforce policy gradient for port selection in slow FAMA. Performance claims rest on Monte-Carlo spectral-efficiency comparisons against external reference schemes under stated channel models. No equation or training step reduces by construction to a self-defined quantity, fitted parameter renamed as prediction, or load-bearing self-citation chain; the methods are defined independently and evaluated numerically without tautological closure.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No mathematical derivations, parameters, or new entities are described in the abstract; the ledger is therefore empty.

pith-pipeline@v0.9.0 · 5403 in / 1031 out tokens · 23513 ms · 2026-05-10T19:25:05.367052+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

GFwd incrementally selects ports by maximizing λ_max(ÃS∪{p}, B̃S∪{p}) and applies swap refinement; Transformer scores ports via MHSA on features fp = [Re(h̃p), Im(h̃p), γ̄p, s̄p, ῑp] then takes top-L.
IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Performance evaluated via Monte-Carlo SE curves under fixed Jakes' model with P=100, L=8, K=10; no sensitivity to alternative geometries or hardware validation.

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Hybrid Multiport Receivers for Slow Fluid Antenna Multiple Access
cs.IT 2026-05 unverdicted novelty 6.0

A fluid-antenna hybrid multiport receiver achieves performance close to full-digital multiport schemes using only 2 RF chains and cuts computational load by over 60 percent in slow multiuser scenarios.

Reference graph

Works this paper leans on

16 extracted references · 1 canonical work pages · cited by 1 Pith paper

[1]

Fluid antenna systems: Redefining reconfigurable wireless communications,

W. K. New, K.-K. Wong, C. Wang, C.-B. Chae, R. Murch, H. Jafarkhani, and Y . Hao, “Fluid antenna systems: Redefining reconfigurable wireless communications,”IEEE J. Sel. Areas Commun., vol. 44, pp. 1013–1044, 2026

2026
[2]

Performance limits of fluid antenna systems,

K. K. Wong, A. Shojaeifard, K.-F. Tong, and Y . Zhang, “Performance limits of fluid antenna systems,”IEEE Commun. Lett., vol. 24, no. 11, pp. 2469–2472, 2020

2020
[3]

Fluid antenna multiple access,

K.-K. Wong and K.-F. Tong, “Fluid antenna multiple access,”IEEE Trans. Wireless Commun., vol. 21, no. 7, pp. 4801–4815, 2022

2022
[4]

Slow fluid antenna multiple access,

K.-K. Wong, K.-F. Tong, Y . Chen, and Y . Zhang, “Slow fluid antenna multiple access,”IEEE Trans. Commun., vol. 71, no. 5, pp. 2831–2846, May 2023

2023
[5]

Compact ultra massive antenna array: A simple open-loop massive connectivity scheme,

K.-K. Wonget al., “Compact ultra massive antenna array: A simple open-loop massive connectivity scheme,”IEEE Trans. Wireless Com- mun., vol. 23, no. 6, pp. 6279–6294, Jun. 2024

2024
[6]

Slow fluid antenna multiple access with multiport receivers,

J. P. Gonz ´alez-Coma and F. J. L ´opez-Mart´ınez, “Slow fluid antenna multiple access with multiport receivers,”IEEE Wireless Commun. Lett., vol. 15, pp. 1280–1284, 2026

2026
[7]

Multi-port selection for FAMA: Massive connectivity with fewer RF chains than users,

H. Hong, K.-K. Wong, X. Zhu, H. Xu, H. Xiao, F. R. Ghadi, and H. Shin, “Multi-Port Selection for FAMA: Massive Connectivity with Fewer RF Chains than Users,”arXiv preprint arXiv:2511.17897, 2025

work page arXiv 2025
[8]

Machine learning-based antenna selection in wireless commu- nications,

J. Joung, “Machine learning-based antenna selection in wireless commu- nications,”IEEE Commun. Surveys Tuts., vol. 23, no. 4, pp. 2371–2388, 2021

2021
[9]

Deep learning enabled slow fluid antenna multiple access,

N. Waqar, K.-K. Wong, K.-F. Tong, A. Sharples, and Y . Zhang, “Deep learning enabled slow fluid antenna multiple access,”IEEE Commun. Lett., vol. 27, no. 3, pp. 861–865, Mar. 2023

2023
[10]

Channel estimation and reconstruction in fluid antenna system: Oversampling is essential,

W. Kiat New, K.-K. Wong, H. Xu, F. Rostami Ghadi, R. Murch, and C.- B. Chae, “Channel estimation and reconstruction in fluid antenna system: Oversampling is essential,”IEEE Trans. Wireless Commun., vol. 24, no. 1, pp. 309–322, 2025

2025
[11]

How much training is required for channel estimation in fluid antenna system?

J.-M. Kang and I.-M. Kim, “How much training is required for channel estimation in fluid antenna system?”IEEE J. Sel. Areas Commun., vol. 44, pp. 1259–1275, 2026

2026
[12]

A new spatial block-correlation model for fluid antenna systems,

P. Ram ´ırez-Espinosaet al., “A new spatial block-correlation model for fluid antenna systems,”IEEE Trans. Wireless Commun., vol. 23, no. 11, pp. 15 829–15 843, Nov. 2024

2024
[13]

Solution of the multiuser downlink beam- forming problem with individual SINR constraints,

M. Schubert and H. Boche, “Solution of the multiuser downlink beam- forming problem with individual SINR constraints,”IEEE Trans. Veh. Technol., vol. 53, no. 1, pp. 18–28, Jan. 2004

2004
[14]

Attention is all you need,

A. Vaswani, N. Shazeer, N. Parmaret al., “Attention is all you need,” in Proc. Adv. Neural Inf. Process. Syst. (NeurIPS), 2017, pp. 5998–6008

2017
[15]

Simple statistical gradient-following algorithms for connectionist reinforcement learning,

R. J. Williams, “Simple statistical gradient-following algorithms for connectionist reinforcement learning,”Mach. Learn., vol. 8, no. 3–4, pp. 229–256, 1992

1992
[16]

G. H. Golub and C. F. V . Loan,Matrix Computations, 4th ed. Johns Hopkins Univ. Press, 2013

2013