There is a trend to deploy neural network on edge devices in recent years. While the mainstream of research often concerns with single edge device processing and edge-cloud two-layer neural network collaborative computing, in this paper, we propose partitioning multi-layer edge network for neural network collaborative computing. With the proposed method, sub-models of neural network are deployed on multi-layer edge devices along the communication path from end users to cloud. Firstly, we propose an optimal path selection method to form a neural network collaborative computing path with lowest communication overhead. Secondly, we establish a time-delay optimization mathematical model to evaluate the effects of different partitioning solutions. To find the optimal partition solution, an ordered elitist genetic algorithm (OEGA) is proposed. The experimental results show that, compared with traditional cloud computing, single-device edge computing and edge-cloud collaborative computing, the proposed multi-layer edge network collaborative computing has a smaller runtime delay with limited bandwidth resources, and because of the pipeline computing characteristics, the proposed method has a better response speed when processing large number of requests. Meanwhile, the OEGA algorithm has better performance than conventional methods, and the optimized partitioning method outperforms other methods like random and evenly partition.

It is necessary to optimize the energy-efficient (EE) performance for millimeter-wave (mmWave) wireless systems, since the power consumption problem becomes increasingly crucial at high frequency bands. In hybrid precoding mmWave systems, the EE-oriented optimized elements include the analog and digital precoders at the transmitter, the analog and digital combiners at the receiver, as well as the corresponding number of opened radio frequency (RF) chains. In contrary to calculate the optimal precoding/combining matrices exhaustively for all possible numbers of RF chains as in the existing works, this paper proposes a two-stage alternative method with nearly optimal sum rate performance and much lower computation complexity. It firstly calculates the optimal number of opened RF chains, and then the nearly optimal precoder and combiner are directly obtained within only one round calculation. The complicated calculations of all the elements in optimization are decoupled by exploring the asymptotical property of the mmWave system with large antenna arrays, as well as the sparsity characteristic of the mmWave fading channel. Simulation results are further provided to verify the validation of the proposed method.