Center for Advanced Intelligence Project

Speaker: Dr. Reinmar Kobler, RIKEN AIP

Title: Selected applications of machine learning in non-invasive brain-computer interfacing and functional neuroimaging

Abstract:
Inherent variability in magnetoencephalographic (MEG) and electroencephalographic (EEG) activity across subjects, recording sessions and tasks limits generalization of brain-computer interfaces (BCI) and biomarker development. A key challenge is to disentangle task-related activity in the presence of non-stationary background activity and artifacts. This talk is divided into two parts. In the first part, we introduce neuroimaging tools which are sensitive to movement trajectory related effects in the EEG and MEG activity and invariant to co-varying eye artifacts. Using these tools, we were able to identifying and characterizing two cortical networks that encode directional and non-directional kinematic information during goal-directed discrete and continuous movements. The second part of the talk will deal with inter-subject and -session transfer learning as well as model interpretation. The covered topics include generative pre-training and discriminative adaptation to aid inter-subject transfer, using Riemannian geometry to train interpretable models that improve generalization to new sessions, and how batch renormalization on the SPD manifold can be used to combine feature learning with Riemannian tangent space classification.