Sparse Optimization Using an MCN for Scattered Signal Suppression

In recent years, there have been reports of using deep learning to optimize the sparse array. In this context, we propose a new network that directly generates an array layout. In the new network, the concept of multimodal is introduced to improve the optimized performance. We processed the original complex information matrix by extracting its real part, imaginary part, and phase. The real and imaginary parts are used as one set of inputs, while the phase is used as another set. Feature extraction is performed separately on the two modalities, followed by data fusion. Simulation results demonstrate that the proposed multimodal control network (MCN) can achieve a classification accuracy of 97.8%, indicating that the proposed multimodal control network has practical application value in sparse array optimization for scattered signal suppression.