Answers are grounded in this paper and cite their sources.
The increasing popularity of Internet of Things (IoT) has led to a large volume of data being generated from these devices. Machine Learning (ML) techniques are frequently used to process such data, enabling IoT systems to predict events and enhance decision-making. ML computation is usually performed in edge devices to improve the real-time responsiveness of the IoT systems. ML demands considerable computational and energy resources, which can be challenging in resource-constrained IoT and edge environments. In this paper, we focus on analyzing the energy costs associated with different stages of the ML pipeline on edge computing hardware. By evaluating the energy usage across different ML steps such as data processing, model training, and inference on diverse hardware configurations, this study aims to improve understanding and provide a basis for energy-efficient ML workflows on the edge. The main contributions of this work include a review of the current state-of-the-art in energy-aware ML for edge computing, an evaluation of energy consumption across different ML pipeline stages under different model parameters, and the derivation of a formula for quantifying the energy cost of an ML instance to support energy-aware decision making. This analysis will guide the development of energy-aware ML solutions that optimize resource usage, enabling sustainable and scalable IoT applications.
No takes yet. Share an insight, caveat, or question.
Sharma et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: