Subspace Diffusion Generative Models

By Bowen Jing, Gabriele Corso, Renato Berlinghieri, Tommi S. Jaakkola. 2022

92 citations recorded by the source corpus.

Score-based models generate samples by mapping noise to data (and vice versa) via a high-dimensional diffusion process. We question whether it is necessary to run this entire process at high dimensionality and incur all the inconveniences thereof. Instead, we restrict the diffusion via projections onto subspaces as the data distribution evolves toward noise. When applied to state-of-the-art models, our framework simultaneously improves sample quality -- reaching an FID of 2.17 on unconditional CIFAR-10 -- and reduces the computational cost of inference for the same number of denoising steps. Our framework is fully compatible with continuous-time diffusion and retains its flexible capabilities, including exact log-likelihoods and controllable generation. Code is available at https://github.com/bjing2016/subspace-diffusion.

Subfields: Computer Vision and Pattern Recognition, Radiology, Nuclear Medicine and Imaging, Statistical and Nonlinear Physics, Generative Models, Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Radiology, Nuclear Medicine and Imaging

In Statistical and Nonlinear Physics

In Generative Models

In Representation & Self-Supervised Learning

Citations in the corpus

Elucidating the Design Space of Diffusion-Based Generative Models (2022, background): “The current high-resolution diffusion models rely either on separate super-resolution steps [17, 35, 39], subspace projection [23], very large networks [9, 48], or hybrid approaches [38, 41, 50] — we believe that our contributions are orthogonal to these extensions.”
Torsional Diffusion for Molecular Conformer Generation (2022, background): “Among these, the most relevant to this work is subspace diffusion [Jing et al., 2022], in which the diffusion is progressively restricted to linear subspaces.”
Diffusion Models: A Comprehensive Survey of Methods and Applications (2022, background): “Furthermore, since the latent space in diffusion models often possesses the same dimensionality as the data space, sampling efficiency is negatively affected and the models may not learn the representation schemes well [106].”
Illuminating protein space with a programmable generative model (2022, background): “Subspace diffusion models [Jing et al., 2022] also consider correlated diffusion, with a particular emphasis on focusing the diffusion to most relevant factors of variation for statistical and computational efficiency.”
Improving Diffusion Model Efficiency Through Patching (2022, methodology): “Our work is most closely related to Subspace Diffusion models [15], who add dimensionalityreducing projections at specified points of diffusion process.”