Generative modelling in latent space – Sander Dieleman

Latent representations for generative models.

Most contemporary generative models of images, sound and video do not operate directly on pixels or waveforms. They consist of two stages : first, a compact, higher-level latent representation is extracted, and then an iterative generative process operates on this representation instead. How does this work, and why is this approach so popular? Generative models that make use of latent representations are everywhere nowadays, so I thought it was high time to dedicate a blog post to them. In what follows, I will talk at length about latents as a plural noun, which is the usual shorthand for late

saved by

• Sebastian J Zhao
• Vincent Cheng
• Darius Kianersi
• Ali Ahmad
• arya
• Clem von Stengel

related reading

• What are Diffusion Models? | Lil'Loglilianweng.github.io
• Musings on typicality – Sander Dielemansander.ai
• Diffusion is spectral autoregression – Sander Dielemansander.ai
• Generating music in the waveform domain – Sander Dielemansander.ai
• The Principles of Diffusion Modelsarxiv.org
• What are Diffusion Models?lilianweng.github.io
• Tips for Training Likelihood Modelsblog.evjang.com
• Learning the integral of a diffusion model – Sander Dielemansander.ai
• [2006.11239] Denoising Diffusion Probabilistic Modelsarxiv.org
• Diffusion language models – Sander Dielemansander.ai
• Diffusion models are autoencoders – Sander Dielemansander.ai
• Difference between AutoEncoder (AE) and Variational AutoEncoder (VAE) | Towards Data Sciencetowardsdatascience.com