Hybrid precoding design based on dual-layer deep-unfolding neural network

G Zhang, X Fu, Q Hu, Y Cai, G Yu - 2021 IEEE 32nd Annual …, 2021 - ieeexplore.ieee.org
2021 IEEE 32nd Annual International Symposium on Personal, Indoor …, 2021ieeexplore.ieee.org
Dual-layer iterative algorithms are generally required when solving resource allocation
problems in wireless communication systems. Specifically, the spectrum efficiency
maximization problem for hybrid precoding architecture is hard to solve by the single-layer
iterative algorithm. The dual-layer penalty dual decomposition (PDD) algorithm has been
proposed to address the problem. Although the PDD algorithm achieves significant
performance, it requires high computational complexity, which hinders its practical …
Dual-layer iterative algorithms are generally required when solving resource allocation problems in wireless communication systems. Specifically, the spectrum efficiency maximization problem for hybrid precoding architecture is hard to solve by the single-layer iterative algorithm. The dual-layer penalty dual decomposition (PDD) algorithm has been proposed to address the problem. Although the PDD algorithm achieves significant performance, it requires high computational complexity, which hinders its practical applications in real-time systems. To address this issue, we first propose a novel framework for deep-unfolding, where a dual-layer deep-unfolding neural network (DLDUNN) is formulated. We then apply the proposed frame-work to solve the spectrum efficiency maximization problem for hybrid precoding architecture. An efficient DLDUNN is designed based on unfolding the iterative PDD algorithm into a layer-wise structure. We also introduce some trainable parameters in place of the high-complexity operations. Simulation results show that the DLDUNN presents the performance of the PDD algorithm with remarkably reduced complexity.
ieeexplore.ieee.org
以上显示的是最相近的搜索结果。 查看全部搜索结果