Generative Modelling With Inverse Heat Dissipation
This work addresses a limitation in image generation for AI and computer vision, but it appears incremental as it builds on existing diffusion models with a novel PDE-based approach.
The authors tackled the problem of diffusion models not explicitly considering image structure by proposing a new diffusion-like model that generates images through stochastically reversing the heat equation, resulting in emergent qualitative properties like disentanglement of overall colour and shape in images.
While diffusion models have shown great success in image generation, their noise-inverting generative process does not explicitly consider the structure of images, such as their inherent multi-scale nature. Inspired by diffusion models and the empirical success of coarse-to-fine modelling, we propose a new diffusion-like model that generates images through stochastically reversing the heat equation, a PDE that locally erases fine-scale information when run over the 2D plane of the image. We interpret the solution of the forward heat equation with constant additive noise as a variational approximation in the diffusion latent variable model. Our new model shows emergent qualitative properties not seen in standard diffusion models, such as disentanglement of overall colour and shape in images. Spectral analysis on natural images highlights connections to diffusion models and reveals an implicit coarse-to-fine inductive bias in them.