publications
publications in reversed chronological order.
2025
-
How Compositional Generalization and Creativity Improve as Diffusion Models are TrainedIn 42nd International Conference on Machine Learning (ICML), 2025 -
Scaling Laws and Representation Learning in Simple Hierarchical Languages: Transformers vs. Convolutional ArchitecturesarXiv preprint, 2025 -
Learning curves theory for hierarchically compositional data with power-law distributed featuresIn 42nd International Conference on Machine Learning (ICML), 2025
2024
-
Towards a theory of how the structure of language is acquired by deep neural networksIn Advances in Neural Information Processing Systems (NeurIPS) 37, 2024 -
How Deep Neural Networks Learn Compositional Data: The Random Hierarchy ModelPhys. Rev. X, 2024 -
2023
-
How deep convolutional neural networks lose spatial information with trainingMach. learn.: sci. technol., 2023 -
What Can Be Learnt With Wide Convolutional Neural Networks?In 40th International Conference on Machine Learning (ICML), 2023
2022
-
Learning sparse features can lead to overfitting in neural networksIn Advances in Neural Information Processing Systems (NeurIPS) 35, 2022
2021
-
Locality defeats the curse of dimensionality in convolutional teacher-student scenariosIn Advances in Neural Information Processing Systems (NeurIPS) 34, 2021