Miika Aittala: Elucidating the Design Space of Diffusion-Based Generative Models

Abstract: We argue that the theory and practice of diffusion-based generative models are currently unnecessarily convoluted and seek to remedy the situation by presenting a design space that clearly separates the concrete design choices. This lets us identify several changes to both the sampling and training processes, as well as preconditioning of the score networks. Together, our improvements yield new state-of-the-art FID of 1.79 for CIFAR-10 in a class-conditional setting and 1.97 in an unconditional setting, with much faster sampling (35 network evaluations per image) than prior designs. To further demonstrate their modular nature, we show that our design changes dramatically improve both the efficiency and quality obtainable with pre-trained score networks from previous work, including improving the FID of a previously trained ImageNet-64 model from 2.07 to near-SOTA 1.55, and after re-training with our proposed improvements to a new SOTA of 1.36.
The paper was presented at NeurIPS 2022, where it received an Outstanding Paper Award. Joint work with Tero Karras, Timo Aila and Samuli Laine.
Bio: Miika Aittala is a Senior Research Scientist at NVIDIA Research, which he joined in 2019. He received his PhD in 2016 from Aalto University, working on capture and rendering of surface material appearance. Prior to his current position, he worked as a postdoctoral researcher at MIT CSAIL and visited Inria Sophia Antipolis. His research interests include neural generative modeling and image processing, and realistic image synthesis in computer graphics.

[@susdoge3767]

this video is criminally underrated, thanks for your insights!

[@user-uu5ml5dc6n]

Really clear explanation of how the diffusion network works !! Thanks

[@luke2642]

Fantastic talk, and a really good paper. Over 400 citations in two years... a lot to sift through! Would love to see a follow up! I need to dive into the guidance next, this just leaves me wondering why we don't use clip to guide the diffusion of a low resolution semantic map first, that captures the structure, meaning, long range patterns easily, maybe even a course depth map too, and then use that to guide the diffusion of a latent, which then gets intelligently upscaled guided by the semantics.

[@fyremaelstrom2896]

Excellent elucidation.

[@zbaker0071]

Intelligent insights.

[@nickjordan6360]

Nice talk