Background overview

The following pages describe our experimental methodology as well as discuss basic concepts relevant to our findings.

Scaled Error

Understanding scaled measures of error

Scale, Shape and Domain Size

Key characteristics of the data that affect the error introduced by a differentially private algorithm.

Exchanging Scale and Epsilon

The scale of the data and the epsilon privacy parameter contribute to error in equivalent ways.

Baseline algorithms

Baseline algorithms help to define the limits of reasonable utility.

Exploring Data Dependence

Data-dependent algorithms perform differently on different datasets.


Glossary of Terms

Dataset Overviews

Algorithm Overviews

Overview of Differential Privacy