/non-submodular-max

Source code for the paper "Guarantees for Greedy Maximization of Non-submodular Functions with Applications" in ICML 2017

Primary LanguageMATLAB

Guarantees for Greedy Maximization of Non-submodular Functions

This repository collects source code for the paper:

"Guarantees for Greedy Maximization of Non-submodular Functions with Applications."

ICML 2017. A. Bian, J. Buhmann, A. Krause and S. Tschiatschek.

Usage:

File Structure:

The subfolders contain source files for the experiments:

  • "Bayesian-A-Optimality": Bayesian experimental design with the A-optimality objective

  • "LP": LPs with combinatorial constraints

  • "Determinantal-Function": Determinantal functions maximization

The subfolder "utilities" contains the utility functions.

One can refer to the README file in each subfolder to run the experiments and tune the setup parameters.

Dependencies:

  • This implementation uses the SDP solver provided by CVX (Version 2.1) and the MATLAB LP solver linprog

  • The code has been tested on Ubuntu 16.04 LTS, 64 bits with MATLAB R2016a. It should work with other OS with little change.

Copyright:

Copyright (2017) [Yatao (An) Bian | ETH Zurich]

Please cite the above paper if you are using this code in your work.

The code may be used free of charge for non-commercial and educational purposes, the only requirement is that this text is preserved within the derivative work. For any other purpose one must contact the authors for permission.