We provide theoretical convergence guarantees for score-based generative models (SGMs) such as denoising diffusion probabilistic models (DDPMs), which constitute the backbone of large-scale real-world generative models such as DALL$\cdot$E 2. Our main result is that, assuming accurate score estimates, such SGMs can efficiently sample from essentially any realistic data distribution. In contrast to prior works, our results (1) hold for an $L^2$-accurate score estimate (rather than $L^\infty$-accurate); (2) do not require restrictive functional inequality conditions that preclude substantial non-log-concavity; (3) scale polynomially in all relevant problem parameters; and (4) match state-of-the-art complexity guarantees for discretization of the Langevin diffusion, provided that the score error is sufficiently small. We view this as strong theoretical justification for the empirical success of SGMs. We also examine SGMs based on the critically damped Langevin diffusion (CLD). Contrary to conventional wisdom, we provide evidence that the use of the CLD does not reduce the complexity of SGMs.
翻译:2. 我们的主要结果是,假设准确的得分估计数,这种SGM可以有效地从任何现实的数据分布中进行抽样;与以前的工作不同,我们的结果(1) 持有价值为2,200美元的准确得分估计数(而不是价值为1美元/美元/美元/美元-准确);(2) 不要求有限制功能不平等条件,排除实质性非log-concavity;(3) 在所有相关问题参数中,规模是多元的;(4) 与最先进的朗埃文分散化的复杂保证相匹配,条件是得分误差足够小;我们认为,这是SGM成功经验的坚实理论依据;我们还根据严重达标的Langevin传播(CLDD)研究SGM。 与传统智慧相反,我们提供的证据是,使用SGM没有降低SG的复杂程度。