10 citations · 25 across the 14 of their papers we have counts for
16 papers
Memorization in image classification: test set behavior and scaling laws
Jeevesh Attri, Florentin Guth
Deep neural networks are capable of both learning generalizing solutions on natural tasks and memorizing arbitrary mappings. We argue that understanding the former requires underst…
Localizing Transfer Between Memorization Tasks
Yimiao Yu, Florentin Guth
A central puzzle in transfer learning is why pre-training on one task can accelerate training or improve performance on another task, and what mechanisms underlie this transfer. In…
Learning Normalized Energy Models for Linear Inverse Problems
Nicolas Zilberstein, Santiago Segarra, Eero Simoncelli +1
Generative diffusion models can provide powerful prior probability models for inverse problems in imaging, but existing implementations suffer from two key limitations: the p…
There Will Be a Scientific Theory of Deep Learning
Jamie Simon, Daniel Kunin, Alexander Atanasov +11
In this paper, we make the case that a scientific theory of deep learning is emerging. By this we mean a theory which characterizes important properties and statistics of the train…
Learning a distance measure from the information-estimation geometry of data
Guy Ohayon, Pierre-Etienne H. Fiquet, Florentin Guth +2
We introduce the Information-Estimation Metric (IEM), a novel form of distance function derived from an underlying continuous probability density over a domain of signals. The IEM…
Learning normalized image densities via dual score matching
Florentin Guth, Zahra Kadkhodaie, Eero P Simoncelli
Learning probability models from data is at the heart of many machine learning endeavors, but is notoriously difficult due to the curse of dimensionality. We introduce a new framew…
Classification-Denoising Networks
Louis Thiry, Florentin Guth
Image classification and denoising suffer from complementary issues of lack of robustness or partially ignoring conditioning information. We argue that they can be alleviated by un…
On the universality of neural encodings in CNNs
Florentin Guth, Brice Ménard
We explore the universality of neural encodings in convolutional neural networks trained on image classification tasks. We develop a procedure to directly compare the learned weigh…
Generalization in diffusion models arises from geometry-adaptive harmonic representations
Zahra Kadkhodaie, Florentin Guth, Eero P. Simoncelli +1
Deep neural networks (DNNs) trained for image denoising are able to generate high-quality samples with score-based reverse diffusion algorithms. These impressive capabilities seem…
Conditionally Strongly Log-Concave Generative Models
Florentin Guth, Etienne Lempereur, Joan Bruna +1
There is a growing gap between the impressive results of deep image generative models and classical algorithms that offer theoretical guarantees. The former suffer from mode collap…
A Rainbow in Deep Network Black Boxes
Florentin Guth, Brice Ménard, Gaspar Rochette +1
A central question in deep learning is to understand the functions learned by deep networks. What is their approximation class? Do the learned weights and representations depend on…
Learning multi-scale local conditional probability models of images
Zahra Kadkhodaie, Florentin Guth, Stéphane Mallat +1
Deep neural networks can learn powerful prior probability models for images, as evidenced by the high-quality generations obtained with recent score-based diffusion methods. But th…
Wavelet Score-Based Generative Modeling
Florentin Guth, Simon Coste, Valentin De Bortoli +1
Score-based generative models (SGMs) synthesize new data samples from Gaussian white noise by running a time-reversed Stochastic Differential Equation (SDE) whose drift coefficient…
Phase Collapse in Neural Networks
Florentin Guth, John Zarka, Stéphane Mallat
Deep convolutional classifiers linearly separate image classes and improve accuracy as depth increases. They progressively reduce the spatial dimension whereas the number of channe…
Separation and Concentration in Deep Networks
John Zarka, Florentin Guth, Stéphane Mallat
Numerical experiments demonstrate that deep neural network classifiers progressively separate class distributions around their mean, achieving linear separability on the training s…
Specification Mining for Smart Contracts with Automatic Abstraction Tuning
Florentin Guth, Valentin Wüstholz, Maria Christakis +1
Smart contracts are programs that manage digital assets according to a certain protocol, expressing for instance the rules of an auction. Understanding the possible behaviors of a…