arXiv:2306.08757 · v2

InfoDiffusion: Representation Learning Using Information Maximizing Diffusion Models

首次提交 2023-06-15 05:48

版本更新 2026-10-26 10:53

雷达观测 2026-10-29 00:36

摘要

While diffusion models excel at generating high-quality samples, their latent variables typically lack semantic meaning and are not suitable for representation learning. Here, we propose InfoDiffusion, an algorithm that augments diffusion models with low-dimensional latent variables that capture high-level factors of variation in the data. InfoDiffusion relies on a learning objective regularized with the mutual information between observed and hidden variables, which is designed to improve latent space quality and to prevent the latents from being ignored by expressive diffusion-based decoders. Empirically, we find that InfoDiffusion learns disentangled and human-interpretable latent representations that are competitive with state-of-the-art generative and contrastive methods, while retaining the high sample quality of diffusion models. Our method enables manipulating the attributes of generated images and has the potential to assist tasks that require exploring a learned latent space to generate quality samples, e.g., generative design.

作者备注：Published in ICML 2023; Code is available on the project page: https://yingheng-wang.com/infodiffusion-page/