arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1907.04068stat.MLcs.LG
收藏

基于生成对抗网络的条件独立性测试

Conditional Independence Testing using Generative Adversarial Networks

University of Cambridge(剑桥大学);The Alan Turing Institute(艾伦·图灵研究所);University of California Los Angeles(加州大学洛杉矶分校)

Alexis Bellot, Mihaela van der Schaar

摘要:

我们考虑检测条件依赖的假设检验问题,重点是高维特征空间。我们的贡献是一个新的检验统计量,基于生成对抗网络的样本,旨在直接近似编码原假设的条件分布,以最大化功率(真阴性率)的方式。我们表明,这样的方法只需要密度近似是可行的,以确保我们控制第一类错误(误报率),特别是,没有假设需要作出的形式的分布或功能依赖。使用合成模拟与高维数据,我们证明了显着的收益,在竞争的方法的权力。此外,我们还说明了使用我们的测试来发现遗传数据中疾病的因果标记。

英文摘要:

We consider the hypothesis testing problem of detecting conditional dependence, with a focus on high-dimensional feature spaces. Our contribution is a new test statistic based on samples from a generative adversarial network designed to approximate directly a conditional distribution that encodes the null hypothesis, in a manner that maximizes power (the rate of true negatives). We show that such an approach requires only that density approximation be viable in order to ensure that we control type I error (the rate of false positives); in particular, no assumptions need to be made on the form of the distributions or feature dependencies. Using synthetic simulations with high-dimensional data we demonstrate significant gains in power over competing methods. In addition, we illustrate the use of our test to discover causal markers of disease in genetic data.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑