Unlocking Guidance for Discrete State-Space Diffusion and Flow Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Nisonoff, Hunter, Xiong, Junhao, Allenspach, Stephan, Listgarten, Jennifer
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910893308116992
author Nisonoff, Hunter
Xiong, Junhao
Allenspach, Stephan
Listgarten, Jennifer
author_facet Nisonoff, Hunter
Xiong, Junhao
Allenspach, Stephan
Listgarten, Jennifer
contents Generative models on discrete state-spaces have a wide range of potential applications, particularly in the domain of natural sciences. In continuous state-spaces, controllable and flexible generation of samples with desired properties has been realized using guidance on diffusion and flow models. However, these guidance approaches are not readily amenable to discrete state-space models. Consequently, we introduce a general and principled method for applying guidance on such models. Our method depends on leveraging continuous-time Markov processes on discrete state-spaces, which unlocks computational tractability for sampling from a desired guided distribution. We demonstrate the utility of our approach, Discrete Guidance, on a range of applications including guided generation of small-molecules, DNA sequences and protein sequences.
format Preprint
id arxiv_https___arxiv_org_abs_2406_01572
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Unlocking Guidance for Discrete State-Space Diffusion and Flow Models
Nisonoff, Hunter
Xiong, Junhao
Allenspach, Stephan
Listgarten, Jennifer
Machine Learning
Generative models on discrete state-spaces have a wide range of potential applications, particularly in the domain of natural sciences. In continuous state-spaces, controllable and flexible generation of samples with desired properties has been realized using guidance on diffusion and flow models. However, these guidance approaches are not readily amenable to discrete state-space models. Consequently, we introduce a general and principled method for applying guidance on such models. Our method depends on leveraging continuous-time Markov processes on discrete state-spaces, which unlocks computational tractability for sampling from a desired guided distribution. We demonstrate the utility of our approach, Discrete Guidance, on a range of applications including guided generation of small-molecules, DNA sequences and protein sequences.
title Unlocking Guidance for Discrete State-Space Diffusion and Flow Models
topic Machine Learning
url https://arxiv.org/abs/2406.01572