Imperial College London
Portrait Picture

ProfessorRuthMisener

Faculty of Engineering, Department of Computing

Professor in Computational Optimisation
 
 
 
//

Contact

 

+44 (0)20 7594 8315r.misener Website CV

 
 
//

Location

 

379Huxley BuildingSouth Kensington Campus

//

Summary

 

Publications

Citation

BibTex format

@inproceedings{Tsay:2021,
author = {Tsay, C and Kronqvist, J and Thebelt, A and Misener, R},
publisher = {arXiv},
title = {Partition-based formulations for mixed-integer optimization of trained ReLU neural networks},
url = {http://arxiv.org/abs/2102.04373v1},
year = {2021}
}
Download

RIS format (EndNote, RefMan)

TY  - CPAPER
AB  - This paper introduces a class of mixed-integer formulations for trained ReLUneural networks. The approach balances model size and tightness by partitioningnode inputs into a number of groups and forming the convex hull over thepartitions via disjunctive programming. At one extreme, one partition per inputrecovers the convex hull of a node, i.e., the tightest possible formulation foreach node. For fewer partitions, we develop smaller relaxations thatapproximate the convex hull, and show that they outperform existingformulations. Specifically, we propose strategies for partitioning variablesbased on theoretical motivations and validate these strategies using extensivecomputational experiments. Furthermore, the proposed scheme complements knownalgorithmic approaches, e.g., optimization-based bound tightening capturesdependencies within a partition.
AU  - Tsay,C
AU  - Kronqvist,J
AU  - Thebelt,A
AU  - Misener,R
PB  - arXiv
PY  - 2021///
TI  - Partition-based formulations for mixed-integer optimization of trained ReLU neural networks
UR  - http://arxiv.org/abs/2102.04373v1
UR  - http://hdl.handle.net/10044/1/86973
ER  -
Download