Demystifying Variational Diffusion Models
This work addresses the need for a clearer understanding of diffusion models for practitioners and new researchers, though it is incremental as it consolidates existing knowledge rather than introducing new methods.
The authors tackled the problem of making diffusion models more accessible by synthesizing a holistic perspective using directed graphical modeling and variational inference, resulting in a narrative that requires fewer prerequisites for understanding.
Despite the growing interest in diffusion models, gaining a deep understanding of the model class remains an elusive endeavour, particularly for the uninitiated in non-equilibrium statistical physics. Thanks to the rapid rate of progress in the field, most existing work on diffusion models focuses on either applications or theoretical contributions. Unfortunately, the theoretical material is often inaccessible to practitioners and new researchers, leading to a risk of superficial understanding in ongoing research. Given that diffusion models are now an indispensable tool, a clear and consolidating perspective on the model class is needed to properly contextualize recent advances in generative modelling and lower the barrier to entry for new researchers. To that end, we revisit predecessors to diffusion models like hierarchical latent variable models and synthesize a holistic perspective using only directed graphical modelling and variational inference principles. The resulting narrative is easier to follow as it imposes fewer prerequisites on the average reader relative to the view from non-equilibrium thermodynamics or stochastic differential equations.