Diffusion is spectral autoregression – Sander Dieleman
A deep dive into spectral analysis of diffusion models of images, revealing how they implicitly perform a form of autoregression in the frequency domain.
A bit of signal processing swiftly reveals that diffusion models and autoregressive models aren’t all that different: diffusion models of images perform approximate autoregression in the frequency domain! This blog post is also available as a Python notebook in Google Colab , with the code used to produce all the plots and animations. Last year, I wrote a blog post describing various different perspectives on diffusion . The idea was to highlight a number of connections between diffusion models and other classes of models and concepts. In recent months, I have given a few talks where I discuss
Explore this link on the map →saved by
- Pranav
- Freeman Jiang
- Dron Hazra
- Thi Vũ
- Jacob G-W
- Sudarsh K
- Yudhister Joel Kumar
- Dhruv Sheth
- Kenneth Sun
- Florent Tavernier
- Ishan Mukherjee
- Ali Ahmad
related reading
- Diffusion is not necessarily Spectral Autoregression | Fabian Falckfabianfalck.com
- What are Diffusion Models? | Lil'Loglilianweng.github.io
- Generating music in the waveform domain – Sander Dielemansander.ai
- Diffusion language models – Sander Dielemansander.ai
- Generative modelling in latent space – Sander Dielemansander.ai
- Diffusion Meets Flow Matchingdiffusionflow.github.io
- Diffusion models from scratchchenyang.co
- Grokking Diffusion Models – Non_Interactive – Software & MLnonint.com
- ⭐️ Diffusion Modelsandrewkchan.dev
- The Principles of Diffusion Modelsarxiv.org
- Diffusion models are autoencoders – Sander Dielemansander.ai
- Diffusion model - Wikipediaen.wikipedia.org