Francesco Cagnetta
6 papers · 2021–2025 · 2 conferences · across top CS/AI conferences
Achievements
Jump to papers ↓+3 more ↓ Show less ↑
π Renaissance Researcher (5) πΊοΈ Taxonomy Completionist (17) π§ Keyword Pioneer π Conference Polyglot (2) π Interdisciplinary Bridge
π
Cross-Pollinator
(14)
π₯
Unstoppable
(5)
β
The Questioner
Conferences
ICML (3)
NIPS (3)
Top co-authors
Keywords
generalization error
(2)
convolutional neural network
(2)
feature learning
(1)
grammar learning
(1)
sparse representation
(1)
self-supervised learning
(1)
kernel regression
(1)
hierarchical structure
(1)
language model
(1)
random feature
(1)
learning curve
(1)
next-token prediction
(1)
probabilistic context-free grammar
(1)
infinite width
(1)
lazy training
(1)
spatial scale adaptation
(1)
kernel methods
(1)
neural network
(1)
grammar structure
(1)
representation learning
(1)
Papers
Learning curves theory for hierarchically compositional data with power-law distributed features
ICML 2025
How Compositional Generalization and Creativity Improve as Diffusion Models are Trained
ICML 2025
Towards a theory of how the structure of language is acquired by deep neural networks
NIPS 2024
What Can Be Learnt With Wide Convolutional Neural Networks?
ICML 2023
Learning sparse features can lead to overfitting in neural networks
NIPS 2022
Locality defeats the curse of dimensionality in convolutional teacher-student scenarios
NIPS 2021