Backpropagation-Free Test-Time Adaptation for Lightweight EEG-Based Brain-Computer Interfaces
The episode discusses a paper proposing Backpropagation-Free Test-Time Adaptation for Lightweight EEG-Based Brain-Computer Interfaces. The method adapts models to new users without updating internal weights, using test-time transformations and a learning-to-rank module to weight predictions. The authors demonstrate its robustness on noisy signals and quantized models, showing it is faster than backpropagation methods.