Score-based generative diffusion with "active" correlated noise sources
This work addresses an incremental improvement in generative modeling for researchers in machine learning, focusing on noise modulation in diffusion models.
The paper tackled the problem of improving generative performance in diffusion models by using temporally correlated noise sources, akin to active matter, in the forward process, and found that the reverse process may exhibit improved generative properties based on numerical and analytical experiments.
Diffusion models exhibit robust generative properties by approximating the underlying distribution of a dataset and synthesizing data by sampling from the approximated distribution. In this work, we explore how the generative performance may be be modulated if noise sources with temporal correlations -- akin to those used in the field of active matter -- are used for the destruction of the data in the forward process. Our numerical and analytical experiments suggest that the corresponding reverse process may exhibit improved generative properties.