dc.contributor.author | Ruan, W | |
dc.contributor.author | Huang, X | |
dc.contributor.author | Kwiatkowska, M | |
dc.date.accessioned | 2020-08-03T15:56:48Z | |
dc.date.issued | 2018-07-19 | |
dc.description.abstract | Verifying correctness of deep neural networks (DNNs) is challenging. We study a generic reachability problem for feed-forward DNNs which, for a given set of inputs to the network and a Lipschitz-continuous function over its outputs, computes the lower and upper bound on the function values. Because the network and the function are Lipschitz continuous, all values in the interval between the lower and upper bound are reachable. We show how to obtain the safety verification problem, the output range analysis problem and a robustness measure by instantiating the reachability problem. We present a novel algorithm based on adaptive nested optimisation to solve the reachability problem. The technique has been implemented and evaluated on a range of DNNs, demonstrating its efficiency, scalability and ability to handle a broader class of networks than state-of-the-art verification approaches. | en_GB |
dc.description.sponsorship | Engineering and Physical Sciences Research Council (EPSRC) | en_GB |
dc.description.sponsorship | NSFC | en_GB |
dc.identifier.citation | Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence (IJCAI-18), 13-19 July 2018, Stockhom, Sweden, pp. 2651 - 2659 | en_GB |
dc.identifier.doi | 10.24963/ijcai.2018/368 | |
dc.identifier.grantnumber | EP/M019918/1 | en_GB |
dc.identifier.grantnumber | 61772232 | en_GB |
dc.identifier.uri | http://hdl.handle.net/10871/122296 | |
dc.language.iso | en | en_GB |
dc.publisher | IJCAI | en_GB |
dc.rights | © 2018 International Joint Conferences on Artificial Intelligence. All right reserved. | en_GB |
dc.subject | Machine Learning: Neural Networks | en_GB |
dc.subject | Agent-based and Multi-agent Systems: Formal Verification, Validation and Synthesis | en_GB |
dc.subject | Computer Vision | en_GB |
dc.subject | Machine Learning: Deep Learning | en_GB |
dc.title | Reachability analysis of deep neural networks with provable guarantees | en_GB |
dc.type | Conference proceedings | en_GB |
dc.date.available | 2020-08-03T15:56:48Z | |
dc.identifier.isbn | 9780999241127 | |
dc.identifier.issn | 1045-0823 | |
dc.description | This is the final version. Available from IJCAI via the DOI in this record | en_GB |
dc.rights.uri | http://www.rioxx.net/licenses/all-rights-reserved | en_GB |
rioxxterms.version | VoR | en_GB |
rioxxterms.licenseref.startdate | 2018-07-19 | |
rioxxterms.type | Conference Paper/Proceeding/Abstract | en_GB |
refterms.dateFCD | 2020-08-03T15:52:38Z | |
refterms.versionFCD | VoR | |
refterms.dateFOA | 2020-08-03T15:56:53Z | |
refterms.panel | B | en_GB |