1 paper · 1 filter
Zheyang Shen, Huihui Wang, Marina Riabiz +1
Diffusion models are typically trained using score matching, a learning objective agnostic to the underlying noising process that guides the model. This paper argues that Markov no…