Laboratoire de Génie Informatique et d’Automatique de l’Artois

This thesis is situated at the crossroads of several themes: predictive uncertainty quantification, feature extraction, and the understanding and conservation of marine ecosystems. While predictive uncertainty quantification, which is a fundamental and topical subject in machine learning, will be the main scientific theme, feature extraction will also play an important role given the nature of the data to be processed.

The successful candidate will deliver teaching in cybersecurity, Windows/Linux systems, and virtualization within the BUT in Networks and Telecommunications at the IUT of Béthune. He/she will also contribute to student supervision, educational projects, and collective academic responsibilities.

In research, the candidate must have a strong scientific background and solid expertise in Artificial Intelligence applied to at least one of the laboratory’s scientific themes: decision-making and information fusion, or optimization of complex systems. He/she will conduct research in AI applied to autonomous robotics, drones, and Industry 4.0 within the LGI2A, with strong involvement in national projects and academic and industrial collaborations.

The objective of this internship is to combine a CNN capable of automatically classifying sleep stages from EEG signals extracted from the Sleep-EDF database with the CD-ENN model, in order to create a DNN that can both predict sleep stages and precisely indicate the level of uncertainty associated with each prediction.