论文标题
GDDIM：广义降级扩散隐式模型
gDDIM: Generalized denoising diffusion implicit models
论文作者
论文摘要
我们的目标是将脱氧扩散模型（DDIM）扩展到一般扩散模型〜（DMS）以外的一般扩散模型。我们从数值的角度研究了DDIM的机理，而不是像原始DDIM中构造非马尔科夫的noising过程。我们发现，在求解相应的随机微分方程时，可以通过使用分数的一些特定近似值来获得DDIM。我们提出了DDIM加速效应的解释，该解释还解释了确定性抽样方案的优势，而不是随机采样方案进行快速采样。在此洞察力的基础上，我们将DDIM扩展到了General DMS，它创造了广义DDIM（GDDIM），并在参数化分数网络时进行了小但微妙的修改。我们在两个非各向异性DMS中验证GDDIM：模糊扩散模型（BDM）和严格抑制的Langevin扩散模型（CLD）。我们观察到BDM的加速度超过20倍。在CLD中，通过速度增强扩散过程的扩散模型，我们的算法在CIFAR10上获得了FID分数为2.26，只有50次数量的得分函数评估〜（NFES），而FID得分为2.86，仅为2.86，只有27个NFE。代码可从https://github.com/qsh-zh/gddim获得
Our goal is to extend the denoising diffusion implicit model (DDIM) to general diffusion models~(DMs) besides isotropic diffusions. Instead of constructing a non-Markov noising process as in the original DDIM, we examine the mechanism of DDIM from a numerical perspective. We discover that the DDIM can be obtained by using some specific approximations of the score when solving the corresponding stochastic differential equation. We present an interpretation of the accelerating effects of DDIM that also explains the advantages of a deterministic sampling scheme over the stochastic one for fast sampling. Building on this insight, we extend DDIM to general DMs, coined generalized DDIM (gDDIM), with a small but delicate modification in parameterizing the score network. We validate gDDIM in two non-isotropic DMs: Blurring diffusion model (BDM) and Critically-damped Langevin diffusion model (CLD). We observe more than 20 times acceleration in BDM. In the CLD, a diffusion model by augmenting the diffusion process with velocity, our algorithm achieves an FID score of 2.26, on CIFAR10, with only 50 number of score function evaluations~(NFEs) and an FID score of 2.86 with only 27 NFEs. Code is available at https://github.com/qsh-zh/gDDIM
