View Item

Domain Machine Learning-Robotics
Domain - extra Neurosciences
Year 2014
Starting sept. 2014
Status Open
Subject The Brain Code
Thesis advisor SEBAG Michèle
Co-advisors Sylvain Chevallier, Université Versailles St Quentin
Laboratory LRI A&O
Collaborations CRICM, La Pitié Salpétrière, Denis Schwartz et Fabrizio de Vico Fallani
Abstract Among the most fascinating applications of Machine Learning are Neuroimaging and
Brain Machine Interfaces, which use the traces of the subject mental activity to respectively
build a model of the brain functional structure, or convert the subject’s brainwave functional
data into the command law of a mechatronic device.

Among the main ML challenges faced by both above domains is the variability of the
brainwave data, among different subjects and among different sessions for a given subject.
Assuming that single subject brainwave data do involve recurrent information patterns, the PhD
aims at building a subject-dependent code exploiting the brainwave functional data regularities.

A non-linear coding approach, based on neural nets (auto-encoders and deep neural nets), will apply to brainwave data the feature design mechanisms at the root of deep learning. The question is
to enforce the interpretability of the non-linear code in order to enable its qualitative evaluation.

Work program
Extra information
Expected funding Research contract
Status of funding Expected
user michele-martine.sebag
Created Thursday 12 of June, 2014 16:06:32 CEST
LastModif Thursday 12 of June, 2014 16:06:32 CEST
Attachments (0)


No attachments for this item

The original document is available at