Learning Representative Port Subsets for Efficient Fluid Antenna Systems

Quan Zhou; Ming Zeng

cs.ITOct 24, 2026

Learning Representative Port Subsets for Efficient Fluid Antenna Systems

Authors: Quan Zhou, Ming Zeng

Organizations: National University of Singapore · Université Laval

Abstract

Fluid antenna systems (FASs) provide many candidate ports, but probing all ports for each channel realization incurs substantial online overhead. We study a complementary offline problem: compress the candidate-port space once and reuse the compressed set across future channel realizations. The method collects optimal ports from sampled training realizations, requiring only one scan of the candidate ports per realization. On a DeepMIMO 7-by-7 FAS with 49 candidate ports, 120 training realizations yield 27 representative ports on average while retaining 99.995% of the channel power gain achieved by exhaustive port selection.

Figures & tables

Symbol	Meaning
$N$	Number of candidate ports in the FAS
$S$	Number of offline training channel realizations
$M$	Number of test channel realizations
$\mathcal{A}$	Generic subset of candidate ports
$\mathcal{A}_{S}$	RPL subset learned from $S$ training realizations
$R$	Number of ports in the RPL subset, $R=\|\mathcal{A}_{S}\|$

TABLE I: Key notation.

Method	Offline cost	Online candidates
RPL	$O(SN)$	$R$
G-BRPL	$O(SN+SRK)^{*}$	$\min\{R,K\}$
F-BRPL	$O(SN)$	$\min\{R,K\}$
Top- $K$ Mean Utility	$O(SN)$	$K$
Empirical Greedy	$O(SNK)$	$K$
Uniform	$O(NK)$	$K$

TABLE II: Offline complexity and retained online candidate size for the subset-selection methods. Here

R=|\mathcal{A}_{S}|

denotes the (random) size of RPL subset , while

K

is the target online budget.

Panel of Figure 3 — Fig. 1: Fixed-budget utility–overhead tradeoff: relative utility loss (upper) and near-optimal coverage (lower) versus the online budget $K$ .

Fig. 2: Relative utility loss versus the common observed-port budget $R$ for RPL and the Chai-style limited-observation LSTM. Each method uses the same number of observed ports.

$S$	$R=\|\mathcal{A}_{S}\|$	RPL	Uniform	Top-K Mean Utility	Empirical Greedy
40	$15.20\!\pm\!2.17$	$(3.07\!\pm\!3.85)\!\times\!10^{-4}$	$(1.54\!\pm\!0.21)\!\times\!10^{-4}$	$(1.61\!\pm\!0.37)\!\times\!10^{-2}$	$(4.57\!\pm\!8.36)\!\times\!10^{-4}$
80	$22.65\!\pm\!2.06$	$(4.26\!\pm\!15.73)\!\times\!10^{-4}$	$(9.50\!\pm\!1.34)\!\times\!10^{-5}$	$(1.13\!\pm\!0.29)\!\times\!10^{-2}$	$(4.36\!\pm\!15.73)\!\times\!10^{-4}$
120	$27.10\!\pm\!2.86$	$(4.64\!\pm\!2.06)\!\times\!10^{-5}$	$(7.43\!\pm\!1.30)\!\times\!10^{-5}$	$(8.77\!\pm\!1.74)\!\times\!10^{-3}$	$(4.88\!\pm\!3.03)\!\times\!10^{-5}$
160	$31.40\!\pm\!2.21$	$(2.62\!\pm\!1.29)\!\times\!10^{-5}$	$(5.87\!\pm\!1.22)\!\times\!10^{-5}$	$(6.74\!\pm\!1.20)\!\times\!10^{-3}$	$(2.38\!\pm\!0.97)\!\times\!10^{-5}$
200	$32.80\!\pm\!2.63$	$(2.35\!\pm\!1.41)\!\times\!10^{-5}$	$(5.44\!\pm\!1.19)\!\times\!10^{-5}$	$(6.08\!\pm\!1.02)\!\times\!10^{-3}$	$(2.30\!\pm\!1.50)\!\times\!10^{-5}$
300	$38.45\!\pm\!1.99$	$(5.35\!\pm\!6.24)\!\times\!10^{-6}$	$(3.83\!\pm\!0.28)\!\times\!10^{-5}$	$(3.93\!\pm\!0.70)\!\times\!10^{-3}$	$(3.52\!\pm\!3.01)\!\times\!10^{-6}$

TABLE III: Offline sample-efficiency sweep: mean learned subset size and relative utility loss (mean

\pm

sample std over

20

repetitions).

Fig. 3: Offline sample efficiency under the deployment distribution. Left: relative utility loss versus training size $S$ , with all methods using an online observation budget $K=R$ ; right: RPL subset cardinality $R$ versus $S$ , with mean $\pm$ one standard deviation over $20$ repetitions.

Wireless Communications

Wireless communications studies how information is sent over radio links and how networks manage signals, spectrum, and resources reliably.

See all 400 papers

Greedy and Transformer-Based Multi-Port Selection for Slow Fluid Antenna Multiple Access
Apr 6, 2026Darian Perez-Adan, Jose P. Gonzalez-Coma, F. Javier Lopez-Martinez +1Wireless Communications Wireless Resource Allocation
Department of Computer Engineering, University of A Coruña, CITIC, A Coruña, Spain · Defense University Center at the Spanish Naval Academy, Marín, Spain · Dept. Signal Theory, Networking and Communications, Research Centre for Information and Communication Technologies (CITIC-UGR), University of Granada, 18071, Granada, Spain
AirPASS: Over-the-Air Federated Learning via Pinching Antenna Systems
Jul 7, 2026Seyed Mohammad Azimi-Abarghouyi, Christopher G. BrintonFederated Learning Aggregation Wireless Communications
Department of Electrical Engineering, Chalmers University of Technology, Gothenburg, Sweden · School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN USA
Beyond LLM-GA: Secure Fluid Antenna Systems with ReEvo-Designed Memetic Algorithm
Oct 7, 2026Hanyong Xu, Zhaolai Dang, Tong ZhangWireless Communications Metaheuristic Optimization
School of Information, Harbin Institute of Technology, Shenzhen, Shenzhen 518055, China · Guangdong Provincial Key Laboratory of Aerospace Communication and Networking Technology, Harbin Institute of Technology, Shenzhen, 518055, China

Apr 6, 2026J/K move · Enter open · S save

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

Department of Computer Engineering, University of A Coruña, CITIC, A Coruña, Spain · Defense University Center at the Spanish Naval Academy, Marín, Spain · Dept. Signal Theory, Networking and Communications, Research Centre for Information and Communication Technologies (CITIC-UGR), University of Granada, 18071, Granada, Spain

Wireless Communications Wireless Resource Allocation

Open paper AskPDF arXiv

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.

Abstract

Wireless Communications

Explore similar work

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

AirPASS: Over-the-Air Federated Learning via Pinching Antenna Systems

Beyond LLM-GA: Secure Fluid Antenna Systems with ReEvo-Designed Memetic Algorithm