Equivariant score-based generative models provably learn distributions with symmetries efficiently

Open in new window