BriefGPT.xyz
May, 2019
数据形态:数据分布的内在距离
Intrinsic Multi-scale Evaluation of Generative Models
HTML
PDF
Anton Tsitsulin, Marina Munkhoeva, Davide Mottin, Panagiotis Karras, Alex Bronstein...
TL;DR
用Gromov-Wasserstein距离的下界,通过对所有数据矩计算,基于内在和多尺度的方法对比数据流形。实验证明,该方法能够有效地识别不同维度未对齐数据的结构,并展示了在评估生成模型质量方面的功效。
Abstract
generative models
are often used to sample high-dimensional data points from a manifold with small intrinsic dimension. Existing techniques for comparing
generative models
focus on global data properties such as
→