An affordable hardware-aware neural architecture search for deploying convolutional neural networks on ultra-low-power computing platforms
2026-06-15 • Machine Learning
Machine LearningArtificial Intelligence
AI summaryⓘ
The authors designed a way to automatically create very small neural networks that can work on tiny, low-power microcontrollers, like those used in small sensors. Their method is light enough to run even on the embedded devices themselves. They tested their approach on three popular tiny vision tasks, and their networks kept accuracy levels similar to the best current models. This helps bring powerful AI to devices with very limited power and computing ability.
Hardware-aware neural architecture searchConvolutional Neural NetworksMicrocontrollersUltra-low-power devicesEmbedded systemsNeural architecture searchTiny computer visionLightweight search procedureClassification accuracy
Authors
Andrea Mattia Garavagno, Edoardo Ragusa, Antonio Frisoli, Paolo Gastaldo
Abstract
Hardware-aware neural architecture search (HW-NAS) allows the integration of Convolutional Neural Networks (CNNs) in microcontrollers devices by automatically designing neural architectures that can fit prearranged hardware constraints. However, state-of-the-art HW-NAS target high-performance microcontrollers, whose power consumption does not meet sensing nodes requirements. This work presents a HW-NAS generating tiny CNNs that can run on ultra-low-power microcontrollers, featuring a lightweight search procedure enabling its execution even on embedded devices. Empirical results on three well-known benchmarks for tiny computer vision proved that the proposed HW-NAS was able to generate tiny CNNs while preserving state-of-the-art classification accuracy.