Unsupervised Learning for Cellular Power Control

This letter applies a feedforward neural network trained in an unsupervised fashion to the problem of optimizing transmit powers cellular wireless systems. Both uplink and downlink are considered, with either centralized or distributed power control. Various objectives entertained, all them such that can be cast convex form. The performance proposed procedure is very satisfactory and, terms computational cost, scalability system dimensionality markedly superior solvers. Moreover, optimization relies on directly measurable channel gains, no need for user location information.