Misclassification risk and uncertainty quantification in deep classifiers
dc.contributor.author | Şensoy, Murat | |
dc.contributor.author | Saleki, Maryam | |
dc.contributor.author | Julier, S. | |
dc.contributor.author | Aydoğan, Reyhan | |
dc.contributor.author | Reid, J. | |
dc.date.accessioned | 2023-01-06T13:48:09Z | |
dc.date.available | 2023-01-06T13:48:09Z | |
dc.date.issued | 2021 | |
dc.identifier.isbn | 978-073814266-1 | |
dc.identifier.issn | 2472-6737 | en_US |
dc.identifier.uri | http://hdl.handle.net/10679/8010 | |
dc.identifier.uri | https://ieeexplore.ieee.org/document/9423198 | |
dc.description.abstract | In this paper, we propose risk-calibrated evidential deep classifiers to reduce the costs associated with classification errors. We use two main approaches. The first is to develop methods to quantify the uncertainty of a classifier’s predictions and reduce the likelihood of acting on erroneous predictions. The second is a novel way to train the classifier such that erroneous classifications are biased towards less risky categories. We combine these two approaches in a principled way. While doing this, we extend evidential deep learning with pignistic probabilities, which are used to quantify uncertainty of classification predictions and model rational decision making under uncertainty.We evaluate the performance of our approach on several image classification tasks. We demonstrate that our approach allows to (i) incorporate misclassification cost while training deep classifiers, (ii) accurately quantify the uncertainty of classification predictions, and (iii) simultaneously learn how to make classification decisions to minimize expected cost of classification errors. | en_US |
dc.description.sponsorship | United States Department of Defense US Army Research Laboratory (ARL) | |
dc.language.iso | eng | en_US |
dc.publisher | IEEE | en_US |
dc.relation.ispartof | 2021 IEEE Winter Conference on Applications of Computer Vision (WACV) | |
dc.rights | restrictedAccess | |
dc.title | Misclassification risk and uncertainty quantification in deep classifiers | en_US |
dc.type | Conference paper | en_US |
dc.publicationstatus | Published | en_US |
dc.contributor.department | Özyeğin University | |
dc.contributor.authorID | (ORCID 0000-0001-8806-4508 & YÖK ID 41438) Şensoy, Murat | |
dc.contributor.authorID | (ORCID 0000-0002-5260-9999 & YÖK ID 145578) Aydoğan, Reyhan | |
dc.contributor.ozuauthor | Şensoy, Murat | |
dc.contributor.ozuauthor | Aydoğan, Reyhan | |
dc.identifier.startpage | 2483 | en_US |
dc.identifier.endpage | 2491 | en_US |
dc.identifier.wos | WOS:000693397600049 | |
dc.identifier.doi | 10.1109/WACV48630.2021.00253 | en_US |
dc.identifier.scopus | SCOPUS:2-s2.0-85116162186 | |
dc.contributor.ozugradstudent | Saleki, Maryam | |
dc.relation.publicationcategory | Conference Paper - International - Institutional Academic Staff and Graduate Student |
Files in this item
Files | Size | Format | View |
---|---|---|---|
There are no files associated with this item. |
This item appears in the following Collection(s)
Share this page