Show simple item record

dc.contributor.authorOmer, Khalid
dc.contributor.authorCaucci, Luca
dc.contributor.authorKupinski, Meredith
dc.date.accessioned2021-03-25T02:08:08Z
dc.date.available2021-03-25T02:08:08Z
dc.date.issued2020-11-01
dc.identifier.citationOmer, K., Caucci, L., & Kupinski, M. (2020). Limitations of CNNs for Approximating the Ideal Observer Despite Quantity of Training Data or Depth of Network. Journal of Imaging Science and Technology.en_US
dc.identifier.issn1062-3701
dc.identifier.doi10.2352/j.imagingsci.technol.2020.64.6.060408
dc.identifier.urihttp://hdl.handle.net/10150/657216
dc.description.abstractThe performance of a convolutional neural network (CNN) on an image texture detection task as a function of linear image processing and the number of training images is investigated. Performance is quantified by the area under (AUC) the receiver operating characteristic (ROC) curve. The Ideal Observer (IO) maximizes AUC but depends on high-dimensional image likelihoods. In many cases, the CNN performance can approximate the IO performance. This work demonstrates counterexamples where a full-rank linear transform degrades the CNN performance below the IO in the limit of large quantities of training data and network layers. A subsequent linear transform changes the images’ correlation structure, improves the AUC, and again demonstrates the CNN dependence on linear processing. Compression strictly decreases or maintains the IO detection performance while compression can increase the CNN performance especially for small quantities of training data. Results indicate an optimal compression ratio for the CNN based on task difficulty, compression method, and number of training images.en_US
dc.language.isoenen_US
dc.publisherSociety for Imaging Science and Technologyen_US
dc.rightsCopyright © Society for Imaging Science and Technology 2020. This article is Open Access under the terms of the Creative Commons CC BY licence.en_US
dc.rights.urihttps://creativecommons.org/licenses/by/4.0/en_US
dc.titleLimitations of CNNs for Approximating the Ideal Observer Despite Quantity of Training Data or Depth of Networken_US
dc.typeArticleen_US
dc.contributor.departmentWyant College of Optical Sciences, University of Arizonaen_US
dc.contributor.departmentDepartment of Medical Imaging, University of Arizonaen_US
dc.identifier.journalJournal of Imaging Science and Technologyen_US
dc.description.noteOpen access articleen_US
dc.description.collectioninformationThis item from the UA Faculty Publications collection is made available by the University of Arizona with support from the University of Arizona Libraries. If you have questions, please contact us at repository@u.library.arizona.edu.en_US
dc.eprint.versionFinal published versionen_US
dc.source.journaltitleJournal of Imaging Science and Technology
dc.source.volume64
dc.source.issue6
dc.source.beginpage60408
dc.source.endpage1
refterms.dateFOA2021-03-25T02:08:21Z


Files in this item

Thumbnail
Name:
s9.pdf
Size:
1.861Mb
Format:
PDF
Description:
Final Published Version

This item appears in the following Collection(s)

Show simple item record

Copyright © Society for Imaging Science and Technology 2020. This article is Open Access under the terms of the Creative Commons CC BY licence.
Except where otherwise noted, this item's license is described as Copyright © Society for Imaging Science and Technology 2020. This article is Open Access under the terms of the Creative Commons CC BY licence.