Diffusion Generative Models in Infinite Dimensions

TL;DR

This paper extends diffusion generative models to operate directly in infinite-dimensional function spaces, enabling more natural modeling of functional data like audio and time series.

Contribution

It develops the theoretical foundation for diffusion models in Hilbert and Sobolev spaces, allowing direct functional data generation and analysis.

Findings

Successful application to synthetic data

Effective unconditional and conditional generation

Theoretical framework for infinite-dimensional diffusion models

Abstract

Diffusion generative models have recently been applied to domains where the available data can be seen as a discretization of an underlying function, such as audio signals or time series. However, these models operate directly on the discretized data, and there are no semantics in the modeling process that relate the observed data to the underlying functional forms. We generalize diffusion models to operate directly in function space by developing the foundational theory for such models in terms of Gaussian measures on Hilbert spaces. A significant benefit of our function space point of view is that it allows us to explicitly specify the space of functions we are working in, leading us to develop methods for diffusion generative modeling in Sobolev spaces. Our approach allows us to perform both unconditional and conditional generation of function-valued data. We demonstrate our methods on several synthetic and real-world benchmarks.