https://arxiv.org/abs/2012.08125
Learning Energy-Based Models by Diffusion Recovery Likelihood (Ruiqi Gao, Yang Song, Ben Poole, Ying Nian Wu, Diederik P. Kingma)
denoising diffusion의 아이디어를 ebm에 결합해서 주어진 노이즈 수준에서 데이터의 확률을 최대화하도록 학습. generative model 빌런들이 또...
[[200619 Denoising Diffusion Probabilistic Models]]
#energy_based_model #ddpm