This is a subgroup meeting of the DIMACS Working Group on Privacy / Confidentiality of Health Data.
In this meeting we will explore problems in combinatorial optimization, graph theory, and the interface between statistics and operations research that arise from issues of data privacy and, more specifically, data de-identification.
This is to be an informal meeting, aimed at involving those with interests in combinatorial optimization, graph theory and the stat/OR interface in working on these problems that have become very important in applications such as health data privacy, government statistical data, and counter-terrorism. The emphasis will be on identifying and working on problems of discrete optimization and on identifying and exploring relevant algorithms. No prior knowledge of the application areas is necessary.
Specific problems of interest to be discussed/examined from the OR perspective include combinatorial structure of the feasible region defined by a partially specified multi-dimensional table or by linked tables; generating extremal points and statistical samples from a feasible region defined by a system of multi-dimensional tabular constraints; and (near)-optimization of (nonlinear) statistical functions over a system of tabular constraints. These problems recently have been approached from the standpoint of the theory of Grobner bases but the intended focus here is on combinatorial and mathematical programming approaches and their computability.