Green machine learning for cellular networks

Abstract

Machine learning (ML) is poised to transform the operation of cellular networks by providing efficient solutions ranging from infrastructure management, reduction of operation cost, and improved end-user experience. ML can address numerous problems facing wireless cellular networks without requiring extensive message passing between central controllers (e.g., eNB) and mobile users. ML can enable cellular networks to utilize offline and distributed solutions. However, much of this progress has been achieved by deploying deep learning algorithms requiring significant computational resources. While it is essential to deploy ML for optimized solutions, decreasing computational complexity and increasing the energy efficiency of ML algorithms are also necessary. This requires a paradigm shift to green machine learning (GML), and this chapter provides an overview of the requirements of GML. In this chapter, the intrinsic challenges of deploying green machine learning-based protocols and their impact on the functionality and design trade-offs will be discussed as well. Additionally, several new performance metrics will be discussed to evaluate the performance of GML algorithms.