Comment by tdesilva

The interesting part is they chose to go with a normalizing flow approach, rather than the industry standard diffusion model approach. Not sure why they chose this direction as I haven’t read the paper yet.