Skip to content

Implementation of partition-based formulations for mixed-integer optimization of trained ReLU neural networks

License

Notifications You must be signed in to change notification settings

cog-imperial/PartitionedFormulations_NN

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 

Repository files navigation

PartitionedFormulations_NN

This repository contains scripts for generating partition-based formulations for trained ReLU neural networks and a few test instances implemented in Gurobi. More details on the method here: https://arxiv.org/abs/2102.04373.

Please cite this work as:

@article{tsay2021partition,
  title   =   {{Partition-based formulations for mixed-integer optimization of trained ReLU neural networks}},
  author  =   {Tsay, Calvin and Kronqvist, Jan and Thebelt, Alexander and Misener, Ruth},
  journal =   {ArXiv},
  volume  =   {2102.04373},
  year    =   {2021}
}

Installing Gurobi

The solver software Gurobi is required to run the examples. Gurobi is a commercial mathematical optimization solver and free of charge for academic research. It is available on Linux, Windows and Mac OS.

Please follow the instructions to obtain a free academic license. Once Gurobi is installed on your system, follow the steps to setup the Python interface gurobipy.

Authors

License

This repository is released under the Apache License 2.0. Please refer to the LICENSE file for details.

Acknowledgements

This work was supported by Engineering & Physical Sciences Research Council (EPSRC) Fellowships to CT and RM (grants EP/T001577/1 and EP/P016871/1), an Imperial College Research Fellowship to CT, a Royal Society Newton International Fellowship (NIF\R1\182194) to JK, a grant by the Swedish Cultural Foundation in Finland to JK, and a PhD studentship funded by BASF to AT.

About

Implementation of partition-based formulations for mixed-integer optimization of trained ReLU neural networks

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages