Lecture DL

Groupe de lecture deep learning

Session 19 : mercredi 13 novembre 2024 , 12h30 (visio zoom - 16 participants)
- Van Thao NGuyen: Masked Autoencoders are scalable vision learners
- Clement Bernard: Accurate structure prediction of biomolecular interactions with AlphaFold 3. - slides
Session 18 : lundi 18 decembre 2023 , 13h00 (visio zoom)'
- Van Thao NGuyen: Data2vec: A general framework for self-supervised learning in speech, vision and language - slides
- Gaspar Roy: Break the Sequential Dependency of LLM Inference Using Lookahead Decoding
Session 17 : mercredi 25 octobre 2023 , 10h30 (visio zoom)'
- Constance Creux: MLPInit: Embarrassingly Simple GNN Training Acceleration with MLP Initialization
- Alice Lacan: 3D Gaussian Splatting for Real-Time Radiance Field Rendering
Session 16 : mercredi 10 mai 2023 , 10h30 (visio zoom)'
- Van Thao Nguyen : Big self-supervised models are strong semi-supervised learners.
- Alex Lence: Taming Transformers for high-resolution image synthesis
Session 15 : vendredi 10 février 2023 , 11h (visio collab)'
- Victoria Bourgeais : GlanceNets: Interpretable, Leak-proof Concept-based Models.
- Farida Zehraoui : The Intriguing Relation Between Counterfactual Explanations and Adversarial Examples
Session 14 : lundi 5 décembre 2022, 15h30 (visio collab)'
- Jean-Daniel Zucker : Beyond neural scaling laws: beating power law scaling via data pruning
- Aurélien Beaude : Multimodal Co-Attention Transformer for Survival Prediction in Gigapixel Whole Slide Images, Chen et al.
Session 13 : lundi 17 octobre 2022, 14h00 (visio collab)'
- Dominique Fourer : TFA-Net: A Deep Learning-Based Time-Frequency Analysis Tool
- Alexandre Heuillet : Méthodes NAS appliquées aux réseaux siamois et à l'apprentissage auto-supervisé
Session 12 : lundi 27 juin 2022, 11h00
- Aurelien Beaude : Hopular - Modern Hopfield Networks for Tabular Data
- Blaise Hanczar : Diffusion models
Session 11 : vendredi 25 mars 2022, 14h00 (visio collab)
- Désiré Sidibe : When Does Self-supervision Improve Few-shot Learning ?
- Alice Lacan : Autoencoder Image Interpolation by Shaping the Latent Space
Session 10 : vendredi 4 fevrier 2022, 13h 30(visio collab)
- Alexandre Heuillet : FBNetV3: Joint Architecture-Recipe Search using Predictor Pretraining
- Ayoub Abraich : Patches Are All You Need?
Session 9 : mercredi 10 novembre 2021, 14h (visio collab)
- Dominique Fourer : Improve the Interpretability of Attention: A Fast, Accurate, and Interpretable High-Resolution Attention Model, slides
- Victoria Bourgeais : Federated continual learning with weighted inter-client transfer, slides
Session 8 : mardi 21 septembre 2021, 11h (visio collab)
- Farida Zehraoui : Attention is not not Explanation
- Louis Becquey : Highly accurate protein structure prediction with AlphaFold
Session 7 : jeudi 3 juin 2021, 15h (visio collab)
- Vincent Vigneron : principal components , minor components and linear neural networks
- Blaise Hanczar : Deep double descent : where bigger models and more data hurt
Session 6 : 29 avril 2021 (visio collab)
- Farida Zehraoui : Attention is not not Explanation
- Tina Issa : Pareto multi-task learning
Session 5 : 5 mars 2021 (visio collab)
- Dominique Fourer : Adversarial poisonning and inverse poisonning against deep learning
- Jean-Christophe Janodet : Distance and Equivalence between Finite State Machines and Recurrent Neural Networks: Computational results
Session 4 : 18 décembre 2020 (visio collab)
- Ayoub Abraich : Causal Shapley Values: Exploiting Causal Knowledge to Explain Individual Predictions of Complex Models
- Louis Becquey : Improved protein structure prediction using potentials from deep learning.(AlphaFold)
Session 3 : 13 novembre 2020 (visio collab)
- Victoria Bourgeais: Graph Neural Net intro & Graph U-nets (Gao, H., & Ji, S. 2019)
- Tai Doan InterFaceGAN: Interpreting the Disentangled Face Representation Learned by GANs (Shen et al. 2020)
Session 2 : 29 septembre 2020 (12p)
- Blaise Hanczar : The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks
- Tina Issa : PROXSGD training structured neural neworks under redularization and constraints
Session 1 : 3 mars 2020 (10p)
- Blaise Hanczar : Learning Representations for Counterfactual Inference
- Farida Zehraoui : Hybrid Predictive Model: When an Interpretable Model Collaborates with a Black-box Model

Intranet