The Annotated Diffusion Model

In this blog post, we'll take a deeper look into Denoising Diffusion Probabilistic Models (also known as DDPMs, diffusion models, score-based generative models or simply autoencoders) as researchers have been able to achieve remarkable results with them for (un)conditional image/audio/video generation. Popular examples (at the time of writing) include GLIDE and DALL-E 2 by OpenAI, Latent Diffusion

[rawwell]

“y construction of the forward qqq process and backward process, each term (except for L0L_0L0​) of the loss is actually the KL divergence between 2 Gaussian distributions which can be written explicitly as an L2-loss with respect to the means!”