playing with Dwork's adaptive holdout and how to use it for a grid-search
Testing Dwork’s resuable holdout on grid-search
./original_methodThis folder contains the a python reimplementation of Dwork’s method that was published as The reusable holdout: Preserving validity in adaptive data analysis and originally hoster here
The notebook linked here walks through an experiment showing thresholdOut being applied in a more typical grid-search supervised learning context.
More discussion is availabe on my blog post here