Files in this item
Believe the HiPe : hierarchical Perturbation for fast, robust, and model-agnostic saliency mapping
Item metadata
dc.contributor.author | Cooper, Jessica | |
dc.contributor.author | Arandjelović, Ognjen | |
dc.contributor.author | Harrison, David J | |
dc.date.accessioned | 2022-05-05T14:30:38Z | |
dc.date.available | 2022-05-05T14:30:38Z | |
dc.date.issued | 2022-09 | |
dc.identifier | 279281529 | |
dc.identifier | 70eb9e9e-4ef2-444f-ac42-a554c154655f | |
dc.identifier | 000797676100006 | |
dc.identifier | 85129531861 | |
dc.identifier.citation | Cooper , J , Arandjelović , O & Harrison , D J 2022 , ' Believe the HiPe : hierarchical Perturbation for fast, robust, and model-agnostic saliency mapping ' , Pattern Recognition , vol. 129 , 108743 . https://doi.org/10.1016/j.patcog.2022.108743 | en |
dc.identifier.issn | 0031-3203 | |
dc.identifier.other | RIS: urn:DBD0362725EC3E79FC02490733F43FC7 | |
dc.identifier.other | ORCID: /0000-0001-9041-9988/work/112711430 | |
dc.identifier.uri | https://hdl.handle.net/10023/25287 | |
dc.description | This work is supported by the Industrial Centre for AI Research in Digital Diagnostics (iCAIRD) which is funded by Innovate UK on behalf of UK Research and Innovation (UKRI) [project number: 104690]. | en |
dc.description.abstract | Understanding the predictions made by Artificial Intelligence (AI) systems is becoming more and more important as deep learning models are used for increasingly complex and high-stakes tasks. Saliency mapping – a popular visual attribution method – is one important tool for this, but existing formulations are limited by either computational cost or architectural constraints. We therefore propose Hierarchical Perturbation, a very fast and completely model-agnostic method for interpreting model predictions with robust saliency maps. Using standard benchmarks and datasets, we show that our saliency maps are of competitive or superior quality to those generated by existing model-agnostic methods – and are over 20× faster to compute. | |
dc.format.extent | 11 | |
dc.format.extent | 3042233 | |
dc.language.iso | eng | |
dc.relation.ispartof | Pattern Recognition | en |
dc.subject | XAI | en |
dc.subject | AI safety | en |
dc.subject | Saliency mapping | en |
dc.subject | Deep learning explanation | en |
dc.subject | Interpretability | en |
dc.subject | Prediction attribution | en |
dc.subject | QA75 Electronic computers. Computer science | en |
dc.subject | QA76 Computer software | en |
dc.subject | DAS | en |
dc.subject.lcc | QA75 | en |
dc.subject.lcc | QA76 | en |
dc.title | Believe the HiPe : hierarchical Perturbation for fast, robust, and model-agnostic saliency mapping | en |
dc.type | Journal article | en |
dc.contributor.sponsor | Innovate UK | en |
dc.contributor.institution | University of St Andrews. School of Computer Science | en |
dc.contributor.institution | University of St Andrews. School of Medicine | en |
dc.contributor.institution | University of St Andrews. Sir James Mackenzie Institute for Early Diagnosis | en |
dc.contributor.institution | University of St Andrews. Cellular Medicine Division | en |
dc.identifier.doi | https://doi.org/10.1016/j.patcog.2022.108743 | |
dc.description.status | Peer reviewed | en |
dc.identifier.grantnumber | TS/S013121/1 | en |
This item appears in the following Collection(s)
Items in the St Andrews Research Repository are protected by copyright, with all rights reserved, unless otherwise indicated.