Profile cover photo
Profile photo
Moritz Hardt

Post has shared content
Excited about my joint work with Cynthia Dwork, Vitaly Feldman, Toni Pitassi, Omer Reingold and Aaron Roth being featured on the Google Research blog. 
Preserving validity in adaptive data analysis

With all data analysis, there is a danger that findings observed in a particular sample do not generalize to the underlying population from which the data were drawn. Adaptive analysis of a data set - where the analyst is informed by data exploration, as well as the results of previous analyses of the data set - can lead to an increased risk of spurious discoveries that are neither prevented nor detected by standard approaches. 

In order to increase the reliability of data driven insights, researchers from Google, Microsoft, IBM, the University of Toronto, Samsung and the University of Pennsylvania have introduced the reusable holdout mechanism, a new methodology for navigating the challenges of adaptivity which allows the analyst to safely validate the results of many adaptively chosen analyses without the need to collect costly fresh data each time. 

Learn more on the Google Research blog, linked below.
Wait while more posts are being loaded