A Phase Transition in Diffusion Models Reveals the Hierarchical Nature of Data
This work provides theoretical insights into the hierarchical nature of data in diffusion models, which is incremental for understanding generative AI mechanisms.
The study investigated how diffusion models capture hierarchical data structure by identifying a phase transition in the backward diffusion process, where high-level features like image class reconstruction drop sharply at a threshold time while low-level features evolve smoothly, validated through experiments on ImageNet models.
Understanding the structure of real data is paramount in advancing modern deep-learning methodologies. Natural data such as images are believed to be composed of features organized in a hierarchical and combinatorial manner, which neural networks capture during learning. Recent advancements show that diffusion models can generate high-quality images, hinting at their ability to capture this underlying compositional structure. We study this phenomenon in a hierarchical generative model of data. We find that the backward diffusion process acting after a time $t$ is governed by a phase transition at some threshold time, where the probability of reconstructing high-level features, like the class of an image, suddenly drops. Instead, the reconstruction of low-level features, such as specific details of an image, evolves smoothly across the whole diffusion process. This result implies that at times beyond the transition, the class has changed, but the generated sample may still be composed of low-level elements of the initial image. We validate these theoretical insights through numerical experiments on class-unconditional ImageNet diffusion models. Our analysis characterizes the relationship between time and scale in diffusion models and puts forward generative models as powerful tools to model combinatorial data properties.