AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Is That the Interpretability Paper?
Unsupervised method is an extremely general method for the task of down scaling. Normalizing flows allow you to get much more complicated and interesting distributions over your latent space, so to match reality better. There's this automatic cycle consistency that comes for free. You have exact cycle consistency because the transformations that you're learning from one domain to the other are invertible.