Researchers Shravel Sharma, Imali Dias, and Kevin Lee have analyzed the energy costs of executing machine learning pipelines on edge computing hardware, deriving a formula to quantify energy consumption across distinct computing stages. Machine learning computation is frequently performed directly on edge devices to improve the real-time responsiveness of Internet of Things systems. However, these machine learning techniques demand considerable computational and energy resources, creating significant operational challenges in resource-constrained edge environments.
The study highlights that the increasing popularity of the Internet of Things has generated large volumes of data from connected devices. Machine learning techniques are regularly deployed to process this generated information, enabling networked systems to predict events and enhance automated decision-making. Because edge computing environments operate under strict resource constraints, managing the heavy computational workload required by these predictive algorithms is critical to maintaining continuous device operation.
To address these constraints, the researchers focused on evaluating the energy costs associated with different stages of the machine learning pipeline. The team examined energy usage across distinct operational steps, specifically analyzing data processing, model training, and inference across diverse hardware configurations. By assessing energy consumption across these pipeline stages under different model parameters, the work provides a foundation to improve technical understanding and support energy-efficient workflows on edge hardware.
A primary contribution of the research is the derivation of a mathematical formula designed to quantify the energy cost of an individual machine learning instance. This formula is developed to support energy-aware decision-making when designing and running computational tasks on edge systems. In addition to deriving the cost formula and evaluating pipeline stages, the authors conducted a comprehensive review of the current state of the art in energy-aware machine learning for edge computing.
According to the authors, the resulting pipeline evaluation and mathematical formula will guide the development of energy-aware machine learning solutions that optimize resource usage. By establishing clear metrics for energy consumption across varying hardware configurations and operational stages, the analysis provides a framework to enable sustainable and scalable Internet of Things applications that operate reliably within resource-constrained edge environments.
Subject of Research: Energy consumption of machine learning pipelines in edge computing
Article Title: The Energy Cost of Machine Learning in Edge Computing
Article References: Sharma, S., Dias, I., & Lee, K. (2026). The Energy Cost of Machine Learning in Edge Computing. International Journal of Networked and Distributed Computing. https://doi.org/10.1007/s44227-026-00139-7
Image Credits: AI Generated
DOI: 10.1007/s44227-026-00139-7
Keywords: edge computing, machine learning, energy consumption, Internet of Things, energy cost formula, data processing, model training, model inference, hardware configurations, energy-aware computing, resource-constrained devices, pipeline stages, model parameters, real-time responsiveness, scalable IoT, sustainable computing, event prediction, decision-making systems
Cite Scienmag News
Teresa Odom. (November 20, 2026). Study Quantifies Machine Learning Energy Costs Across Edge Computing Pipeline Stages. Scienmag. https://scienmag.com/study-quantifies-machine-learning-energy-costs-across-edge-computing-pipeline-stages/
Teresa Odom. "Study Quantifies Machine Learning Energy Costs Across Edge Computing Pipeline Stages." Scienmag, 20 November 2026, https://scienmag.com/study-quantifies-machine-learning-energy-costs-across-edge-computing-pipeline-stages/. Accessed 20 November 2026.
Teresa Odom. "Study Quantifies Machine Learning Energy Costs Across Edge Computing Pipeline Stages." Scienmag. November 20, 2026. https://scienmag.com/study-quantifies-machine-learning-energy-costs-across-edge-computing-pipeline-stages/








