BioTrain: Sub-MB, Sub-50mW On-Device Fine-Tuning for Edge-AI on Biosignals

2026-04-14

AI Summary

This work addresses the performance degradation of biosignal models on edge devices caused by cross-user and cross-session variability. The authors propose BioTrain, a framework that enables full-network online fine-tuning directly on milliwatt-scale microcontrollers—such as GAP9—under stringent constraints of sub-megabyte memory and power consumption below 50 mW. By introducing a custom memory allocator, optimizing network topology, and devising an on-chip training strategy, BioTrain supports efficient backpropagation even with batch normalization layers, thereby preserving both privacy and reliability. Experimental results on EEG/EOG tasks demonstrate that BioTrain improves calibration accuracy for new users by up to 35% over non-adaptive baselines and outperforms last-layer-only fine-tuning by approximately 7%, while maintaining a minimal memory footprint of only 0.67 MB.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Cognitive Modeling & Cognitive Systems: Adaptive Behavior
• Computer Vision: Biometrics, Face, Gesture & Pose

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Economics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments

Abstract

Biosignals exhibit substantial cross-subject and cross-session variability, inducing severe domain shifts that degrade post-deployment performance for small, edge-oriented AI models. On-device adaptation is therefore essential to both preserve user privacy and ensure system reliability. However, existing sub-100 mW MCU-based wearable platforms can only support shallow or sparse adaptation schemes due to the prohibitive memory footprint and computational cost of full backpropagation (BP). In this paper, we propose BioTrain, a framework enabling full-network fine-tuning of state-of-the-art biosignal models under milliwatt-scale power and sub-megabyte memory constraints. We validate BioTrain using both offline and on-device benchmarks on EEG and EOG datasets, covering Day-1 new-subject calibration and longitudinal adaptation to signal drift. Experimental results show that full-network fine-tuning achieves accuracy improvements of up to 35% over non-adapted baselines and outperforms last-layer updates by approximately 7% during new-subject calibration. On the GAP9 MCU platform, BioTrain enables efficient on-device training throughput of 17 samples/s for EEG and 85 samples/s for EOG models within a power envelope below 50 mW. In addition, BioTrain's efficient memory allocator and network topology optimization enable the use of a large batch size, reducing peak memory usage. For fully on-chip BP on GAP9, BioTrain reduces the memory footprint by 8.1x, from 5.4 MB to 0.67 MB, compared to conventional full-network fine-tuning using batch normalization with batch size 8.

Problem

• biosignals
• domain shift
• on-device adaptation
• edge-AI
• fine-tuning

Innovation

• on-device fine-tuning
• ultra-low-power AI
• biosignal adaptation
• memory-efficient backpropagation
• edge AI