The field of wireless communication is witnessing significant advancements with the integration of reconfigurable intelligent surfaces (RIS) in next-generation networks. Recent developments focus on optimizing RIS configurations to enhance spectral efficiency, minimize detection errors, and improve overall system performance. Researchers are exploring innovative approaches, including complex-valued neural networks and unsupervised learning-based methods, to efficiently allocate RIS elements and optimize phase configurations. These advancements have the potential to overcome traditional limitations in wireless communication, such as interference and blockage, and enable more flexible and powerful wave-domain control. Noteworthy papers in this area include:
- A study on deep complex-valued neural-network modeling and optimization of stacked intelligent surfaces, which demonstrates improved throughput and error performance in Rician channels.
- A proposal for unsupervised learning-based element resource allocation for RIS in mmWave networks, which reduces computational overhead and improves system throughput.
- An investigation on beyond diagonal IRS aided OFDM, which maximizes achievable rate under frequency-dependent reflection.