Generative Lecture 6: Diffusion Models
This notes introduces diffusion models briefly, including its forward process, backward process and DDPM algorithm.
そして扉がひらくよ——

学习笔记与生活文章,按发布时间排列。一页笔记,一点灵感,一些值得珍藏的日常。
This notes introduces diffusion models briefly, including its forward process, backward process and DDPM algorithm.
This note introduces GANs through distribution matching and adversarial training, then explains Wasserstein distance, WGAN, and the use of weight clipping and gradient penalties.
The note introduces autoregressive models, covering their training, evaluation, decoding strategies, RNN and Transformer implementations, and computational complexity.
This lecture covers geometric transformations, feature matching with RANSAC, and image blending using Laplacian pyramids and Poisson editing to create seamless panoramas.
月亮