Program

Monday April 17

08h30 - 09h00     Welcome coffee and registration

09h00 - 10h30    Basics in deep learning 1
                            Content: basic NN for classification/regression, train and test sets, metrics, over and underfitting, etc, Perceptron and multi-layer perceptron, stochastic gradient descent, learning rate, logistic regression, activation function, regularization (L1/L2/dropout/early stopping), etc
                           Christian Desrosiers, ETS Montréal

11h00 - 12h30     Basics in deep learning 2
                            Content: Weights initialization, forward and backward propagation, batch size, convolution neural nets (CNN), feature maps, pooling, pretraining and transfer learning, applications
                           Christian Desrosiers, ETS Montréal

12h30                   Lunch (Domus)

14h00 - 15h30      Advanced concepts in deep learning 1
                             Content: Common CNN architectures for classification (VGGNet, ResNet, ...) and localization (FasterRCNN, Yolo) and segmentation (encoder-Decoder, U-Net, ENet, ...)
                              Michaël Sdika, Creatis Lyon

16h00 - 18h00     Hands-on/ Challenge presentation
                               Introduction to MONAI
                               Thibault Pelletier, Kitware

After                     Welcome cocktail (Oxxo)

Tuesday April 18

09h00 - 10h30     Generative, auto-encoders and adversarial methods for medical imaging 
                            Content: Autoencoders, variational autoencoders, GANs, CycleGAN and their training
                            Olivier Bernard, Creatis Lyon and Christian Desrosiers, ETS Montreal

11h00 - 12h30     Weakly supervised deep learning
                            Weakly supervised segmentation, constrained CNN losses, semantic segmentation, semi-supervised learning
                            José Dolz and Christian Desrosiers, ETS Montréal

12h30                   Lunch (Domus)

14h00 - 15h30     Hands-on session 1 (classification)
                            Content: Classification from machine learning to deep learning and introduction to explainability

16h00 - 18h00     Hands-on session 1 - (classification and Challenge)

After                     Poster dinner evening

Wednesday April 19

09h00 - 10h30     Advanced concepts in deep learning 2
                            Content: Few-shot learning
                            José Dolz, ETS Montréal

11h00 - 12h30    Privacy in Machine Learning: From Centralized to Federated Approaches for Medical Data                           
                             Content: Domain adaptation/shift, privacy protection and federated learning, adversarial learning, common pitfalls, incomplete data, etc.
                            Antoine Boutet, Inria Lyon, Carole Frindel, Creatis, Lyon

12h30                   Lunch (Domus)

14h00 - 15h30     Hands-on session 2 (segmentation)
                            Content: MRI Muscle segmentation, U-Net based

16h00 - 18h00     Hands-on session 2 - (segmentation and Challenge), Introduction to PlugAI

After                      Free Lyon visit

Thursday April 20

09h00 - 10h30     Advanced concepts in deep learning 3
                           Content: Attention, transformers and related architectures (ViT, CCT, ...)
                            Nicolas Thome, Sorbonne University, Paris

11h00 - 12h30     Round table: Frugal and sustainable AI - Stefan Duffner, Kamel Guerda, Denis Trystram
                            Animation: Paul de Brem

12h30                  Lunch (Domus)

14h00 - 15h30     Hands-on session 3 (variational autoencoder) Content: Auto-encoders, convolutional auto-encoders, variational auto-encoders, latent spaces

16h00 - 18h00     Hands-on session 3 - (variational autoencoder and Challenge)

After                     On boat Gala dinner

Friday April 21

09h00 - 10h30     Geometric deep learning - Examples on brain surfaces
                            Content: Spectral coordinates and representation, spectral deep learning, brain surface matching and parcellation
                            Hervé Lombaert, ETS Montréal

11h00 - 12h30     Deep learning for image reconstruction
                            Content: Plug-and-play and unrolled methods for image reconstruction problems (e.g.,CT, MR, PET)
                            Nicolas Ducros, Creatis Lyon

12h30                   Lunch (Domus)

14h00 - 15h30     Hands-on session 4 (deep learning for image reconstruction)

16h00 - 18h00     Hands-on session 4 - (deep learning for image reconstruction and Challenge)

Departure

restricted area