作者
Yann LeCun
Deep Learning / Vision
Toward Next-Generation Artificial Intelligence: Catalyzing the NeuroAI Revolution
Comments White paper, 10 pages + 8 pages of references, 1 figures
Augmented Language Models: a Survey
Joint Embedding Predictive Architectures Focus on Slow Features
Comments 4 pages (3 figures) short paper for SSL Theory and Practice workshop at NeurIPS 2022. Code is available at NeurIPS_2022" target="_blank" rel="noopener">https://github.com/vladisai/JEPA_SSL_NeurIPS_2022
Coarse-to-Fine Vision-Language Pre-training with Fusion in the Backbone
Comments NeurIPS 2022. Project Website: https://ashkamath.github.io/FIBER_page
Unsupervised Learning of Structured Representations via Closed-Loop Transcription
Comments 17 pages
Compact and Optimal Deep Learning with Recurrent Parameter Generators
Journal ref WACV 2023
projUNN: efficient method for training deep networks with unitary matrices
VICRegL: Self-Supervised Learning of Local Visual Features
Comments Accepted at NeurIPS 2022
Joint Embedding Self-Supervised Learning in the Kernel Regime
Sparse Coding with Multi-Layer Decoders using Variance Regularization
What Do We Maximize in Self-Supervised Learning?
TiCo: Transformation Invariance and Covariance Contrast for Self-Supervised Visual Representation Learning
Masked Siamese ConvNets
Contrastive and Non-Contrastive Self-Supervised Learning Recover Global and Local Spectral Embedding Methods
Pre-Train Your Loss: Easy Bayesian Transfer Learning with Informative Priors
Comments Code available at https://github.com/hsouri/BayesianTransferLearning
Understanding Dimensional Collapse in Contrastive Self-supervised Learning
Comments In Proceedings of the 10th International Conference on Learning Representations (ICLR) 2022
Journal ref ICLR 2022
Separating the World and Ego Models for Self-Driving
Comments 8 pages main content, 14 with references and appendix. 5 figures in total. Submitted and accepted to ICLR 2022 workshop on Generalizable Policy Learning in the Physical World (https://ai-workshops.github.io/generalizable-policy-learning-in-the-physical-world/)
The Effects of Regularization and Data Augmentation are Class Dependent
A Data-Augmentation Is Worth A Thousand Samples: Exact Quantification From Analytical Augmented Sample Moments
VICReg: Variance-Invariance-Covariance Regularization for Self-Supervised Learning
Comments Accepted at ICLR 2022
Neural Manifold Clustering and Embedding
Learning in High Dimension Always Amounts to Extrapolation
MDETR -- Modulated Detection for End-to-End Multi-Modal Understanding
Barlow Twins: Self-Supervised Learning via Redundancy Reduction
Comments 13 pages, 6 figures, to appear at ICML 2021
Inspirational Adversarial Image Generation
Journal ref TIP 2021
Implicit Rank-Minimizing Autoencoder
Unsupervised Image Matching and Object Discovery as Optimization
Comments Accepted to CVPR 2019
Learning about an exponential amount of conditional distributions
Comments 8 pages, 7 figures
