Learning and Adaptive Data Analysis via Maximal Leakage

Learning and Adaptive Data Analysis via Maximal Leakage
复制标题

通过最大泄漏学习和自适应数据分析

DOI:
10.1109/itw44776.2019.8989057
复制
发表时间:
2019
期刊:
2019 IEEE Information Theory Workshop (ITW)
影响因子:
--
通讯作者:
Ibrahim Issa
Ibrahim Issa
中科院分区:
--
文献类型:
--
作者:
A. Esposito;M. Gastpar;Ibrahim Issa

文献摘要

参考文献

被引文献

相似文献

人们越来越有兴趣研究学习算法的泛化误差与信息度量之间的关系。在这项工作中,我们推广了一个使用最大泄漏的结果,并探索了这个界限如何在不同的场景中应用。其主要应用在于对泛化误差进行定界。我们没有分析预期误差,而是提供了一个集中度不等。在这项工作中,我们不需要假设$\sigma$-subGaussian性,并展示了我们的结果如何用于在自适应场景中恢复经典界的推广(例如,c-敏感函数的McDiarmid不等式,通过显著水平的错误发现错误控制等)。
There has been growing interest in studying connections between generalization error of learning algorithms and information measures. In this work, we generalize a result that employs the maximal leakage, a measure of leakage of information, and explore how this bound can be applied in different scenarios. The main application can be found in bounding the generalization error. Rather than analyzing the expected error, we provide a concentration inequality. In this work, we do not require the assumption of $\sigma $-sub gaussianity and show how our results can be used to retrieve a generalization of the classical bounds in adaptive scenarios (e.g., McDiarmid’s inequality for c–sensitive functions, false discovery error control via significance level, etc.
DOI: 10.1109/tit.2019.2962804
发表时间: 2020-03-01
影响因子: 2.5
作者:
Issa, Ibrahim;Wagner, Aaron B.;Kamath, Sudeep
通讯作者: Kamath, Sudeep