Hierarchically branched diffusion models leverage dataset structure for class-conditional generation

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Tseng, Alex M., Shen, Max, Biancalani, Tommaso, Scalia, Gabriele
Formato: Preprint
Publicado: 2022
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910315060396032
author Tseng, Alex M.
Shen, Max
Biancalani, Tommaso
Scalia, Gabriele
author_facet Tseng, Alex M.
Shen, Max
Biancalani, Tommaso
Scalia, Gabriele
contents Class-labeled datasets, particularly those common in scientific domains, are rife with internal structure, yet current class-conditional diffusion models ignore these relationships and implicitly diffuse on all classes in a flat fashion. To leverage this structure, we propose hierarchically branched diffusion models as a novel framework for class-conditional generation. Branched diffusion models rely on the same diffusion process as traditional models, but learn reverse diffusion separately for each branch of a hierarchy. We highlight several advantages of branched diffusion models over the current state-of-the-art methods for class-conditional diffusion, including extension to novel classes in a continual-learning setting, a more sophisticated form of analogy-based conditional generation (i.e. transmutation), and a novel interpretability into the generation process. We extensively evaluate branched diffusion models on several benchmark and large real-world scientific datasets spanning many data modalities.
format Preprint
id arxiv_https___arxiv_org_abs_2212_10777
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle Hierarchically branched diffusion models leverage dataset structure for class-conditional generation
Tseng, Alex M.
Shen, Max
Biancalani, Tommaso
Scalia, Gabriele
Machine Learning
Artificial Intelligence
Class-labeled datasets, particularly those common in scientific domains, are rife with internal structure, yet current class-conditional diffusion models ignore these relationships and implicitly diffuse on all classes in a flat fashion. To leverage this structure, we propose hierarchically branched diffusion models as a novel framework for class-conditional generation. Branched diffusion models rely on the same diffusion process as traditional models, but learn reverse diffusion separately for each branch of a hierarchy. We highlight several advantages of branched diffusion models over the current state-of-the-art methods for class-conditional diffusion, including extension to novel classes in a continual-learning setting, a more sophisticated form of analogy-based conditional generation (i.e. transmutation), and a novel interpretability into the generation process. We extensively evaluate branched diffusion models on several benchmark and large real-world scientific datasets spanning many data modalities.
title Hierarchically branched diffusion models leverage dataset structure for class-conditional generation
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2212.10777