Leveraging Edge Machine Learning for Energy-Efficient Communication in IoT Networks for Carbon-Neutrality

In the rapidly evolving landscape of computer-related technologies, the synergy between artificial intelligence (AI) and the Internet of Things (IoT) has garnered increasing interest from both industry and academia. This burgeoning interest has catalyzed the development of applications at the network edge, which have now achieved considerable scale. As a vital and emergent field of research, machine learning at the network edge intertwines two pivotal themes: wireless communication and machine learning. This research area, termed edge machine learning, endeavors to harness vast quantities of mobile data from edge devices to train machine learning models. A primary challenge addressed in this research is the efficient allocation of limited communication resources amidst the data abundance at the network edge. This includes the strategic management of radio resources, with a particular focus on evaluating and utilizing data importance for optimal radio resource management and allocation. For instance, in wireless communications, the signal-to-noise ratio is often considered a critical metric for assessing data importance. This paper summarizes various methodologies and theories from related works, proposing a novel scheduling algorithm predicated on signal-to-noise ratio considerations. A simulation test, leveraging batch training characteristics of convolutional neural networks, is conducted to validate the proposed approach.