Paper-Conference

Generalization in diffusion models arises from geometry-adaptive harmonic representations

We show that diffusion models can enter a generalization regime where the generated images are independent of the samples in the training set by relying on inductive biases towards geometric regularity in images.

Zahra Kadkhodaie, Florentin Guth, Eero P Simoncelli, Stéphane Mallat

Phase collapse in neural networks

We show that deep networks rely on phase collapses to classify images, rather than sparse coding as assumed in previous works, and introduce a simplified architecture with minimal learning while maintaining high accuracy.

Florentin Guth, John Zarka, Stéphane Mallat

Phase collapse in neural networks