Imperial College London
Portrait Picture

ProfessorAlessioLomuscio

Faculty of Engineering, Department of Computing

Professor of Safe Artificial Intelligence
 
 
 
//

Contact

 

+44 (0)20 7594 8414a.lomuscio Website

 
 
//

Location

 

Imperial-XTranslation & Innovation Hub BuildingWhite City Campus

//

Summary

 

Publications

Citation

BibTex format

@inproceedings{Batten:2021:ijcai.2021/301,
author = {Batten, B and Kouvaros, P and Lomuscio, A and Zheng, Y},
doi = {ijcai.2021/301},
pages = {2184--2190},
publisher = {IJCAI},
title = {Efficient neural network verification via layer-based semidefinite relaxations and linear cuts},
url = {http://dx.doi.org/10.24963/ijcai.2021/301},
year = {2021}
}
Download

RIS format (EndNote, RefMan)

TY  - CPAPER
AB  - We  introduce  an  efficient  and  tight  layer-based semidefinite  relaxation  for  verifying  local  robust-ness of neural networks.   The improved tightness is the result of the combination between semidefinite relaxations and linear cuts.  We obtain a computationally efficient method by decomposing the semidefinite formulation into layer wise constraints. By  leveraging  on  chordal  graph  decompositions, we  show  that  the  formulation  here  presented  is provably tighter than current approaches.  Experiments on a set of benchmark networks show that the  approach  here  proposed  enables  the  verification of more instances compared to other relaxation methods. The results also demonstrate that the SDP relaxation here proposed is one order of magnitude faster than previous SDP methods.
AU  - Batten,B
AU  - Kouvaros,P
AU  - Lomuscio,A
AU  - Zheng,Y
DO  - ijcai.2021/301
EP  - 2190
PB  - IJCAI
PY  - 2021///
SP  - 2184
TI  - Efficient neural network verification via layer-based semidefinite relaxations and linear cuts
UR  - http://dx.doi.org/10.24963/ijcai.2021/301
UR  - https://www.ijcai.org/proceedings/2021/301
UR  - http://hdl.handle.net/10044/1/89706
ER  -
Download