Deep Learning-Aided Tabu Search Detection for Large MIMO Systems

The proposed FS-Net detection scheme is a DNN-aided symbol-detection algorithm for MIMO systems. Unlike the prior DetNet and ScNet schemes, the input vector of the FS-Net is optimized such that a reduced number of connections is required in each layer. Furthermore, the correlation between the input and output signals is considered in the loss function of the FS-Net, allowing it to be trained better with faster convergence compared to DetNet and ScNet.