This paper introduces UR2M, a novel Uncertainty and Resource-aware event detection framework for microcontrollers (MCUs). The framework aims to overcome the challenge of unreliable predictions caused by distribution shifts between training and testing data in traditional machine learning techniques. UR2M is designed to provide accurate event detection and reliable uncertainty estimation in wearable event detection (WED) applications. The implementation optimizes system efficiency and offers substantial improvements in terms of inference speed, energy saving, memory usage, and uncertainty quantification performance. The paper suggests that UR2M can be deployed on a wide range of MCUs, significantly expanding the possibilities for real-time and reliable WED applications.
Publication date: 15 Feb 2024
Project Page: Not provided
Paper: https://arxiv.org/pdf/2402.09264
