FEMBA on the Edge: Physiologically-Aware Pre-Training, Quantization, and Deployment of a Bidirectional Mamba EEG Foundation Model on an Ultra-Low Power Microcontroller

Abstract

This work enables continuous, long-term neuro-monitoring on wearable devices by overcoming the computational bottlenecks of Transformer-based EEG foundation models and the quantization challenges inherent to state-space models. FEMBA, a bidirectional Mamba architecture pre-trained on over 21,000 hours of EEG, is combined with physiologically-aware pre-training, quantization, and end-to-end deployment on an ultra-low-power microcontroller.

Thorir Mar Ingolfsson

Postdoctoral Researcher

Making AI run on microwatts: I build foundation models for EEG, ECG, and other biosignals, and compress them until they run in real time on wearable devices, not in the cloud.