Invertible Neural Networks

Glow is a type of generative model based on normalizing flows, designed for efficient and exact likelihood estimation and high-quality image synthesis. Developed by researchers at OpenAI, Glow offers invertible transformations enabling both data generation and density estimation.