Privay Tools collaborator Adam Smith will offer a course on adaptive data analysis at Boston University this fall. Auditors from other universities are welcome to attend.
CS 591: Adaptive Data Analysis, Algorithmic Stability and Privacy
Times: Mondays and Wednesdays, 4:30-5:45pm
(Lectures start this Wednesday, September 6.)
Location: MCS Building B25, Boston University
(111 Cummington Mall, Boston)
Description: Adaptive data analysis is the increasingly common
practice by which insights gathered from data are used to inform
further analysis of the same data sets. This is common practice both
in machine learning, and in scientific research, in which data-sets
are shared and re-used across multiple studies. Standard theory
assumes that the analysis to be performed on a data set is selected
independently of the data set. Unfortunately, when the set of analyses
run is itself a function of the data, much of this theory becomes
invalid. The resulting disconnect has been blamed as one of the causes
of the crisis of reproducibility in empirical science.
This course will look at recently developed approaches to this
problem. We will see approaches stemming from the literature on
"differential privacy", approaches rooted in measuring leaked
information, and approaches coming from more standard statistical
The course is aimed at graduate students in computer science,
statistics and electrical engineering. The prerequisites are a solid
background in probability, and general "mathematical maturity"
(comfort reading and writing definitions, theorems and proofs). The
course will involve reading and reviewing research papers, pencil and
paper assignments, and some programming problems.
Course materials will be posted here: