What Exactly Are Diffusion Models?
Introduction
Full credit goes to the original author, linked below. All blog posts were reposted either with permission of the author, or by anonymous submission by SAIRC members like yourself.
An intuition-first explainer that starts from physical diffusion — perfume spreading through a room, sugar dissolving in water — and builds up to how diffusion models gradually noise data and learn to reverse the process, drawing a helpful contrast with the encoder of a variational autoencoder along the way.