Overview of findings
These are four of the research findings that emerged from our empirical investigation. Each is supported with an interactive visualization on the pages linked below. Please see the
background discussion for details on experimental methodology and key concepts underlying the empirical findings.
Is there one algorithm that outperforms the rest, across diverse input settings?
Finding:
No single algorithm offers uniformly low error. At small scales, data-dependent algorithms dominate; at large scales data-independent algorithms dominate.
Are different data-dependent algorithms exploiting the same properties of datasets?
Finding:
Algorithm error varies significantly with dataset shape and algorithms differ on the dataset shapes on which they perform well.
When do data-dependent algorithms offer improved error rates?
Finding:
Data-dependence can offer significant improvements in error, at smaller scales or lower epsilon values, but some data-dependent algorithms fail to offer benefits at larger scales or higher epsilons.
Do sophisticated algorithms always outperform simple baseline methods?
Finding:
Many algorithms are beaten by the IDENTITY baseline at large scales, in both 1D and 2D. At low scales, many algorithms result in error rates that are comparable to, or worse than, the Uniform baseline.