The paper proposes a robust methodology to evaluate the performance and computational efficiency of non-parametric two-sample tests for validating high-dimensional generative models in scientific applications.
The study focuses on tests built from univariate integral probability measures: the sliced Wasserstein distance, the mean of the Kolmogorov-Smirnov statistics, and a novel sliced Kolmogorov-Smirnov statistic. These metrics can be evaluated in parallel, allowing for fast and reliable estimates of their distribution under the null hypothesis.
The authors also compare these metrics with the recently proposed unbiased Fréchet Gaussian Distance and the unbiased quadratic Maximum Mean Discrepancy, computed with a quartic polynomial kernel.
The proposed tests are evaluated on various distributions, including correlated Gaussians, mixtures of Gaussians in 5, 20, and 100 dimensions, and a particle physics dataset of gluon jets from the JetNet dataset, considering both jet- and particle-level features.
The results demonstrate that one-dimensional-based tests provide a level of sensitivity comparable to other multivariate metrics, but with significantly lower computational cost, making them ideal for evaluating generative models in high-dimensional settings. The methodology offers an efficient, standardized tool for model comparison and can serve as a benchmark for more advanced tests, including machine-learning-based approaches.
toiselle kielelle
lähdeaineistosta
arxiv.org
Tärkeimmät oivallukset
by Samuele Gros... klo arxiv.org 09-26-2024
https://arxiv.org/pdf/2409.16336.pdfSyvällisempiä Kysymyksiä