This paper proposes a deep neural network (DNN) based method to solve the multicell power control problem that maximizes the sum rate subject to per-user rate constraints. The basic idea is to employ a two-DNN concatenating network structure, where the second DNN associated with a randomization processing is designed to guarantee the per-user rate constraints via supervised learning, given which the first DNN is trained to directly maximize the sum rate by unsupervised learning. Simulation results demonstrate that the proposed method can achieve better performance with low complexity compared to existing deep learning and numerical optimization methods.
Constrained learning for Multicell Power Control
2021-04-01
2273762 byte
Conference paper
Electronic Resource
English
Controlled-Shape Multicell Inflatable Structures
Online Contents | 1996
Shear flows in multicell sandwich sections
NTRS | 1948
|Shear flows in multicell sandwich sections
TIBKAT | 1948
|