ABSTRACT
The Internet of Things (IoT) is a key enabler for many future wireless applications, from manufacturing to healthcare. The IoT interconnects many objects (or devices) that perform complex tasks (e.g., data collection, service optimization, data transmission). The advanced information and communications technologies (ICT) surrounding our smart world, including virtual and augmented reality, autonomous and unmanned aerial vehicles, smart grids, to name a few, are equipped with various sensors and actuators that generate a large amount of IoT data, requiring energy-efficient data retrieval, networking, computation, and caching. On the one hand, many IoT devices operate with limited energy storage due to their restricted battery capacity. On the other hand, with machine learning (ML), significant energy is involved in the storage, the forwarding, and the building of ML models. Therefore, interest in green ML techniques has been growing with the aim to provide energy-efficient ML architectures and solutions for IoT applications. The use of ML in IoT networks with the integration of emerging techniques such as edge computing and softwarization can help provide energy-efficient IoT solutions. This chapter provides an overview of the application of green ML in the IoT domain. We give a comprehensive survey that highlights the use of green ML for IoT and describes various green ML solutions for IoT.
