Differentially Private Learning of Geometric Concepts

We present differentially private efficient algorithms for learning union of polygons in the plane (which are not necessarily convex). Our algorithms achieve (alpha, beta)-PAC learning and (epsilon, delta)-differential privacy using a sample of size (O) over tilde( 1/(alpha epsilon)k log d), where the domain is [d] x [d] and CIE k is the number of edges in the union of polygons.