Programme

L'ensemble de la jounée aura lieu dans le Batiment Gustave Ferrié, Département Génie Electrique. 

Cours magistral -  

Ce cours introduira les principaux concepts (modèles, stratégie d'apprentissage et données) pour comprendre comment les réseaux de neurones fonctionnent. Il détaillera en particulier la modélisation d'un neurone artificiel (poids, biais et fonction d'activation) à partir duquel sont élaborés les modèles profonds. Ensuite il présentera comment la stratégie d'apprentissage basée sur l'algorithme du gradient stochastique permet d'entraîner un réseau de neurones pour différentes applications. L'ensemble des notions présentées pourront être manipulées de manière interactive grâce à la plateforme CooLearning (développée par le laboratoire https://github.com/CooLearning/playground). La fin du cours portera sur des aspects pratiques de l’entraînement et du test des réseaux de neurones (séparation des données, stratégies d'évaluation, métriques...), ainsi qu'une présentation des applications qui seront vues l'après-midi lors des travaux pratiques.

Durant le temps de midi, autour d'un buffet, les participants pourront partager leurs expériences et leurs questions concernant la mise en oeuvre des méthodes d'apprentissage automatique et les réseaux de neurones. Les sujets méthodologiques ainsi que les aspects les pratiques jusqu'à la programmation seront partagés entre les participants.

L'après-midi sera consacrée à des travaux pratiques. À partir d'un Jupyter notebook les participants seront amenés à entrainer et tester plusieurs réseaux de neurones vus le matin sur une tâche de classification d'images médicales et non médicales. Odyssée Merveille, Emmanuel Roux, François Varray et Olivier Bernard animeront cette séance.

Les intervenants

Odyssée Merveille received the engineering degree from the Institut Supérieur de BioSciences and the master degree in Signal and Image in Medicine from Paris-Est in 2013 and the PhD degree in computer science from Paris-Est University in 2016. She is currently an Associate Professor at University of Lyon / CREATIS laboratory. Her scientific interests include image filtering and segmentation, variational models and deep learning for medical imaging.

Emmanuel Roux is Associate Professor at the Université Lyon 1 and the CREATIS lab in Lyon (France) since 2019. In 2016 he received his Ph.D. in acoustics and information engineering from the University of Lyon (France) and the University of Florence (Italy). His current research concerns (deep) machine learning for medical imaging, with particular interests on 3-D ultrasound imaging (2-D transducer optimization and novel beamforming techniques), volumetric medical image segmentation (e.g., pathologic lungs CT or lesion quantification in MRI), weakly-supervised anomaly detection (e.g., Doppler signals) and interactive learning (model interpretability and interactive annotations).

Rémi Emonet is Associate Professor (Maître de Conférence) at University Jean-Monnet and is leading of the Machine Learning project at Laboratoire Hubert Curien, in Saint Étienne.
He got a Ph.D. from the Grenoble university working at Inria, and spent some years, at Idiap research institute, Switzerland, working on probablistic models for unsupervised activity modeling in videos.
His current research and contributions focus on transfer learning, deep representation learning and anomaly detection.
He likes to manipulate Bayesian approaches and to try to derive meaningful guarantees for Machine Learning algorithms.

François Varray  received the Diploma in engineering and the master’s degree in image and signal processing from the Ecole des Mines de Saint-Etienne, Saint-Étienne, France, in 2008, and the Ph.D. degree in nonlinear ultrasound simulation from the Université de Lyon 1, Lyon, France, in 2011. His Ph.D. research was realized in co-agreement between the CREATIS, Lyon, and the MSD Laboratory, Florence, Italy, where he has been an Associate Professor since 2013. His current research interests include the nonlinear ultrasound propagation simulation, nonlinear image simulation, multiresolution motion estimation, cardiac imaging, and photoacoustic imaging. Recently, he participated to the developpement of 3-D+t ultrafast acquisition system.

Olivier Bernard received the Electrical Engineering degree and the Ph.D. degree from the University of Lyon (INSA), France, in 2003 and 2006, respectively. In 2007, he was a Postdoctoral Fellow with the Biomedical Imaging Group at the Federal Polytechnic Institute of Lausanne, EPFL, Switzerland. He is currently a Professor with the University of Lyon (INSA) and the CREATIS laboratory, France, where he is also head of the Myriad research team. His current research interests include image analysis through deep learning techniques with applications in cardiovascular imaging and blood flow imaging.