Conference - Deep Learning : from theory to applications

Tuesday 4 September
(13h Reception opens)
14h Keynote 1: Stéphane Mallat (ENS Paris / Collège de France) "Unsupervised learning: from maximum entropy to deep generative networks"
15h Session on Maths and Signal
- François Malgouyres (Institut de Mathématiques de Toulouse, Univ. Paul Sabatier): “Multilinear compressive sensing and an application to convolutional linear networks
- Nicolas Courty (Univ. Bretagne Sud / IRISA):”Optimal Transport and Domain Adaptation”
- Vincent Gripon (IMT Atlantique): “Convolutional Neural Networks for Signals on Graphs"
- Michel Sebag (LRI): "Searching Causal Models"
16h30 Coffee break
17h Panel on scientific challenges, organized by Gérard Biau (Sorbonne Université)
18h END

Wednesday 5 September
(8h30 Reception opens)
9h Keynote 2: Pierre Vandergheynst (EPFL) Geometric deep learning and applications
10h Coffee break
10h30 Session on Mathematical Tools
- Johannes Schmidt-Hieber (Leiden University): "Statistical theory for deep ReLU networks"
- Gerardo Rubino (Inria): "Machine Learning with Random Neural Networks''
- Antoine Chatalic (Univ. Rennes 1): "Compressive learning with random moments"
- Erwan Le Merrer (Tecnicolor): “Decision boundaries & security related questions
12h Lunch
13h30 Keynote 3: Cordelia Schmid (Inria) "Automatic understanding of the visual world"
14h30 Session on Audio
- Romain Hennequin (Deezer): “Audio-based metric learning for artist disambiguation in large music catalogs
- Fabian-Robert Stöter (Inria & LIRMM / Univ Montpellier): “Deep Learning for Music Unmixing”
- Sanjeel Parekh (Telecom ParisTech / Technicolor): “Representation learning for audio-visual events
- Laurent Girin (Univ. Grenoble Alpes, Grenoble-INP, GIPSA-lab): “Deep learning for speech enhancement
16h Coffee Break
16h30 Keynote 4: François Pachet (Spotify) "Controling generative models for music generation: from Markov chains to deep networks"
17h30 END

Thursday 6 September
(8h30 Reception opens)
9h Keynote 5: Matthieu Cord (Sorbonne Université) "Visual Question Answering: A new task for vision and language understanding"
10h Coffee Break
10h30 Session on Images
- Martin Engilberge (Sorbonne Université / Technicolor): “Deep semantic-visual embedding with localization
- Elisa Fromont (Univ. Rennes 1): "Deep learning from imbalanced data"
- Gilles Puy (Technicolor): “A flexible convolutional solver with application to photorealistic style transfer
- Thierry Dumas (Inria): “Deep learning for intra picture prediction
12h Lunch
13h30 Keynote 6: Olivier Pietquin (Google Brain): “Deep reinforcement learning with demonstrations”
14h30 Session on Language
- Julien Perez (Naver Labs Europe): “Machine Reading, recent advances
- Loic Barrault (LIUM, Le Mans Université): "Multimodal machine translation”
15h20 Farewell coffee