Picture for Matthieu Wyart

Matthieu Wyart

Probing the Latent Hierarchical Structure of Data via Diffusion Models

Add code
Oct 17, 2024
Figure 1 for Probing the Latent Hierarchical Structure of Data via Diffusion Models
Figure 2 for Probing the Latent Hierarchical Structure of Data via Diffusion Models
Figure 3 for Probing the Latent Hierarchical Structure of Data via Diffusion Models
Figure 4 for Probing the Latent Hierarchical Structure of Data via Diffusion Models
Viaarxiv icon

Could ChatGPT get an Engineering Degree? Evaluating Higher Education Vulnerability to AI Assistants

Add code
Aug 07, 2024
Viaarxiv icon

Towards a theory of how the structure of language is acquired by deep neural networks

Add code
May 28, 2024
Viaarxiv icon

How Deep Networks Learn Sparse and Hierarchical Data: the Sparse Random Hierarchy Model

Add code
Apr 16, 2024
Viaarxiv icon

A Phase Transition in Diffusion Models Reveals the Hierarchical Nature of Data

Add code
Mar 04, 2024
Viaarxiv icon

On the different regimes of Stochastic Gradient Descent

Add code
Sep 21, 2023
Viaarxiv icon

How Deep Neural Networks Learn Compositional Data: The Random Hierarchy Model

Add code
Jul 31, 2023
Viaarxiv icon

Dissecting the Effects of SGD Noise in Distinct Regimes of Deep Learning

Add code
Jan 31, 2023
Viaarxiv icon

How deep convolutional neural networks lose spatial information with training

Add code
Oct 04, 2022
Figure 1 for How deep convolutional neural networks lose spatial information with training
Figure 2 for How deep convolutional neural networks lose spatial information with training
Figure 3 for How deep convolutional neural networks lose spatial information with training
Figure 4 for How deep convolutional neural networks lose spatial information with training
Viaarxiv icon

How Wide Convolutional Neural Networks Learn Hierarchical Tasks

Add code
Aug 01, 2022
Figure 1 for How Wide Convolutional Neural Networks Learn Hierarchical Tasks
Figure 2 for How Wide Convolutional Neural Networks Learn Hierarchical Tasks
Viaarxiv icon