Andisheh Amrollahi
Efficiently Learning Fourier Sparse Set Functions
Andisheh Amrollahi, Amir Zandieh, Michael Kapralov, Andreas Krause
Learning set functions is a key challenge arising in many domains, ranging from sketching graphs to black-box optimization with discrete parameters. In this paper we consider the problem of efficiently learning set functions that are defined over a ground set of size n and that are sparse (say k-sparse) in the Fourier domain. This is a wide class, that includes graph and hypergraph cut functions, decision trees and more.
Efficiently Learning Fourier Sparse Set Functions
Andisheh Amrollahi, Amir Zandieh, Michael Kapralov, Andreas Krause
Neural Information Processing Systems http://nips.cc/