EvoBrain: Continual Learning of EEG Foundation Models Across Heterogeneous BCI Tasks
Researchers have introduced EvoBrain, a novel framework designed for continual learning in EEG foundation models. This approach tackles the challenge of adapting models across various brain-computer interface tasks without requiring task-specific fine-tuning for each new application. EvoBrain employs techniques like Neuro-Spectral Task Normalization and Response-Affinity Distillation to manage the balance between learning new information and retaining old knowledge, aiming to create a unified system for brain decoding. AI
IMPACT Enables more scalable and efficient brain-computer interfaces by allowing models to learn new tasks without forgetting previous ones.