Unsupervised Learning for Brain Computer Interfaces (BCI)
Thesis advisor
SEBAG Michèle
Co-advisors
BCIs measure the electro-magnetic currents induced by neuronal activity, at the surface of the skull, and use them to control electronic or mechanical devices (wheelchair, videogame).
BCIs training is usually handled as a supervised statistical learning problem - see the BCI Challenges.
It faces two bottlenecks: i/ the brain is a black box; ii/ which is evolving rapidly.