Oshawott

Summary
The paper presents a novel approach for energy-efficient downlink transmission using active stacked intelligent metasurfaces (ASIM). It addresses the challenge of optimizing both beamforming at the base station and ASIM configurations, leveraging alternative optimization and Bayesian techniques. The work is well-structured, with clear problem formulation, transformation, and solution methods.

Mathematical/empirical assessment
The paper provides a solid mathematical foundation, including channel modeling, power consumption analysis, and EE maximization. The use of epigraph and Lagrangian dual transformations, along with successive convex approximation and Bayesian optimization, is well-motivated and appropriate for the non-convex nature of the problem. The numerical results demonstrate the effectiveness of the proposed method compared to passive SIM and heuristic benchmarks.

Strengths
The paper effectively addresses a complex and relevant problem in wireless communication systems. The proposed framework is comprehensive, considering practical aspects such as amplification constraints, power consumption, and near-field coupling. The use of Bayesian optimization for ASIM configuration is innovative and well-explained. The empirical results are convincing and provide useful insights into the design of active metasurface-assisted networks.

Concerns
While the paper is technically sound, the complexity of the proposed algorithm may pose challenges for real-time implementation. Additionally, the paper could benefit from a more detailed discussion of the computational complexity and convergence behavior of the Bayesian optimization component. A comparison with other optimization techniques, such as gradient-based methods, might also provide further insight into the advantages of the proposed approach.

Final decision
Weak accept