Among numerous emerging technologies targeting future wireless communication scenarios and demands, Reconfigurable Intelligent Surface stand out as a research hotspot due to their advantages in controlling the wireless communication environment with lower cost and energy consumption. This article uses double-layer RIS to replace traditional components as multiple-in multiple-out transmitters, and utilizes the diffraction between RIS layers to achieve signal level error control encoding. Under this model, this article simplifies the Alternating Direction Method of Multipliers algorithm using the proximal gradient method, and optimizes the algorithm by introducing parameters and training with deep learning. This reduces the complexity of the algorithm to a certain extent and improves its performance.