Score-based generative diffusion with 'active' correlated noise sources

Autor:	Lamtyugina, Alexandra, Behera, Agnish Kumar, Nandy, Aditya, Floyd, Carlos, Vaikuntanathan, Suriyanarayanan
Rok vydání:	2024
Předmět:	Computer Science - Machine Learning Condensed Matter - Disordered Systems and Neural Networks
Druh dokumentu:	Working Paper
Popis:	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. Comment: 18 pages, 11 figures
Databáze:	arXiv
Externí odkaz:	http://arxiv.org/abs/2411.07233 Zobrazit plný text záznamu View this record from Arxiv