Complexity-Aware ANN-Based Energy Efficiency Maximization

This work deals with the use of artificial neural networks for energy efficiency optimization. Unlike previous works, it addresses the question of how frequently should the neural network be re-trained in order to optimize the long-term energy efficiency of a wireless network. This question is motivated by the fundamental trade-off between frequently updating the configuration of the neural network in response to changes in the propagation channel statistics, and the energy consumption of the training process. In order to shed light on this tradeoff, this work develops energy consumption models that quantity the energy consumption due to the training and use of a neural network. Moreover, the long-term energy efficiency performance of power control based on neural networks is compared to state-of-the-art methods based only on the use of optimization theory.