Adaptive Data Analysis with Correlated Observations

Aryeh Kontorovich, Menachem Sadigurschi*, Uri Stemmer

*Corresponding author for this work

Research output: Contribution to journalConference articlepeer-review

6 Scopus citations

Abstract

The vast majority of the work on adaptive data analysis focuses on the case where the samples in the dataset are independent. Several approaches and tools have been successfully applied in this context, such as differential privacy, max-information, compression arguments, and more. The situation is far less well-understood without the independence assumption. We embark on a systematic study of the possibilities of adaptive data analysis with correlated observations. First, we show that, in some cases, differential privacy guarantees generalization even when there are dependencies within the sample, which we quantify using a notion we call Gibbs-dependence. We complement this result with a tight negative example. Second, we show that the connection between transcript-compression and adaptive data analysis can be extended to the non-iid setting.

Original languageEnglish
Pages (from-to)11483-11498
Number of pages16
JournalProceedings of Machine Learning Research
Volume162
StatePublished - 2022
Event39th International Conference on Machine Learning, ICML 2022 - Baltimore, United States
Duration: 17 Jul 202223 Jul 2022

Funding

FundersFunder number
Ben-Gurion University of the Negev
Lynn and William Frankel Center for Computer Science at Ben-Gurion University
Blavatnik Family Foundation
Israel Science Foundation1871/19

    Fingerprint

    Dive into the research topics of 'Adaptive Data Analysis with Correlated Observations'. Together they form a unique fingerprint.

    Cite this