Low-Complexity Learning-Based Max-Min Fairness for SIM-Assisted MISO Communications

In this letter, we investigate the user fairness of stacked intelligent metasurface (SIM)-assisted multiple-input single-output (MISO) systems. Specifically, we focus on maximizing the minimum spectral efficiency (SE) for downlink communications by joint precoding design, i.e., the base station power control and SIM wave-based beamforming. Due to the non-convexity and high complexity of the formulated problem, we propose an efficient low-complexity SIM continual learning-based framework (LC-SIMCL). Specifically, we embed phase coordination into a product manifold structure, reducing the search space from exponential to linear complexity while maintaining physical feasibility and achieving robust performance. Finally, the numerical results validate the effectiveness of the proposed joint precoding framework in promoting user fairness and highlight its millisecond-level optimization speed.