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.

Location: Otaniemi CS Building room T5 (+ zoom)
Time: monday 2:15pm, February 20th