Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
The paper tackles the computationally expensive port-selection problem in fluid antenna multiple access (FAMA) systems. They introduce GFwd+S, a greedy forward-selection method with swap refinement, achieving superior spectral efficiency compared to existing methods. Additionally, they propose a Transformer-based neural network, trained with imitation learning and refined with Reinforce, that approximates the performance of GFwd+S with reduced computational complexity.
Achieve near-optimal spectral efficiency in fluid antenna multiple access with a Transformer-based approach that slashes computational cost.
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.