GeMM

Background:

A brain-computer interface (BCI) based on EEG is a system for sending commands to a machine without using the motor system: only brain signals are used. A major problem with current BCI systems is the need for calibration before each use of the system.

Student contributions:

The goal of this internship is to design, implement, and test a new transfer learning method in order to build a general machine learning model (GeMM) that can be used "out of the box," that is, without any training. The results will be tested on a large dataset, including a number of open-access BCI databases available at GIPSA-lab and elsewhere. The student will learn how to handle, pre-process, and process a large amount of EEG-based BCI data. He will investigate domain adaptation methods, which will require the acquisition of skills in linear algebra and signal processing methods. Finally, the student will learn how to apply machine learning methods in the field of BCI, both shallow (Elastic-Net Logistic Regression) and deep (Riemannian neural networks).