论文标题
高斯化流动
Gaussianization Flows
论文作者
论文摘要
迭代高斯化是一个定点迭代程序，可以将任何连续的随机矢量转化为高斯。基于迭代高斯化，我们提出了一种新型的归一化流量模型，该模型既可以有效地计算样本生成样品的可能性和有效的反转。我们证明，在某些规律性条件下，这些称为高斯化流量的模型（称为高斯化流量）是连续概率分布的通用近似值。由于这种保证的表现力，他们可以捕获多模式目标分布而不会损害样品产生的效率。在实验上，我们表明，与其他有效的可逆流模型（如真实的NVP，Glow和Ffjord）相比，高斯流量在几个表格数据集上的性能更好或可比性。特别是，高斯流动更容易初始化，相对于训练数据的不同转换，表现出更好的鲁棒性，并在小型训练集上更好地推广。
Iterative Gaussianization is a fixed-point iteration procedure that can transform any continuous random vector into a Gaussian one. Based on iterative Gaussianization, we propose a new type of normalizing flow model that enables both efficient computation of likelihoods and efficient inversion for sample generation. We demonstrate that these models, named Gaussianization flows, are universal approximators for continuous probability distributions under some regularity conditions. Because of this guaranteed expressivity, they can capture multimodal target distributions without compromising the efficiency of sample generation. Experimentally, we show that Gaussianization flows achieve better or comparable performance on several tabular datasets compared to other efficiently invertible flow models such as Real NVP, Glow and FFJORD. In particular, Gaussianization flows are easier to initialize, demonstrate better robustness with respect to different transformations of the training data, and generalize better on small training sets.
