However, compared to single-step generative models like GANs, VAEs, and normalizing flows, the iterative generation procedure of diffusion models typically requires 10–2000 times more compute, causing slow inference and limiting their real time applications.
문제 제기 : GAN, VAE와 같은 모델들과 비교해 10-2000번 더 연산해야함. 그래서 제한이 있고 느림.
Our objective is to create generative models that facilitate efficient, single-step generation without sacrificing important advantages of iterative refinement. These advantages include the ability to trade-off compute for sample quality when necessary, as well as the capability to perform zeroshot data editing tasks.
저자는 더 효율적으로 만듦. 반복적인 refinment의 중요한 장점 희생 없이 single-step generation을 얻음. 이 장점은 필요할때 퀄리티와 연산량을 조절할 뿐만 아니라 editing tasks에 zeroshot이 가능함.
As illustrated in Fig. 1, we build on top of the probability flow (PF) ordinary differential equation (ODE) in continuous-time diffusion models (Song et al., 2021), whose trajectories smoothly transition the data distribution into a tractable noise distribution.
Fig.1을 보면, probability flow(PF) 맨위에 설계하고 continuous-time diffusio model ODE를 두어 trajectories를 부드럽게 바꿈.
A notable property of our model is self-consistency: points on the same trajectory map to the same initial point. We therefore refer to such models as consistency models.
중요한 요소는 self-consistency: 똑같은 initial point에 똑같은 trajectory map으로 point함. 저자는 consistency model로 일컬음.
Importantly, by chaining the outputs of consistency models at multiple time steps, we can improve sample quality and perform zero-shot data editing at the cost of more compute, similar to what iterative refinement enables for diffusion models.
multiple time step에 consistency model의 ouputs에 의해 저자는 퀄리티는 높이고 더많은 연산에 zeroshot data editing이 가능함.
<aside> 💡 저자가 Yang Song이라 알더라도 수학이 포함되면 자세히 정리해놓음.
</aside>
Diffusion models start by diffusing pdatapxq with a stochastic differential equation (SDE)
where t P r0, Ts, T ą 0 is a fixed constant, µ(¨, ¨) and g(¨) are the drift and diffusion coefficients respectively, and twtutPr0,Ts denotes the standard Brownian motion
µp와 g는 drift와 diffusion coefficients임. 그리고 w는 brownian motion임.
We denote the distribution of xt as ptpxq and as a result p0pxq ” pdatapxq. A remarkable property of this SDE is the existence of an ordinary differential equation (ODE), dubbed the Probability Flow (PF) ODE by Song et al. (2021), whose solution trajectories sampled at t are distributed according to pt(x):
