Adaptive convolution kernel for artificial neural networks

Basit Öğe Kaydı
dc.authorid0000-0002-8649-6013
dc.authorid0000-0002-5639-0648
dc.contributor.authorTek, Faik Borayen_US
dc.contributor.authorÇam, İlkeren_US
dc.contributor.authorKarlı, Denizen_US
dc.date.accessioned2021-01-29T08:43:58Z
dc.date.available2021-01-29T08:43:58Z
dc.date.issued2021-02
dc.departmentIşık Üniversitesi, Mühendislik Fakültesi, Bilgisayar Mühendisliği Bölümüen_US
dc.departmentIşık University, Faculty of Engineering, Department of Computer Engineeringen_US
dc.departmentIşık Üniversitesi, Fen Edebiyat Fakültesi, Matematik Bölümüen_US
dc.departmentIşık University, Faculty of Arts and Sciences, Department of Mathematicsen_US
dc.descriptionThis work was supported by The Scientific and Technological Research Council of Turkey programme (TUBITAK-1001 no: 118E722), Isik University BAP programme, Turkey (no: 16A202), and NVIDIA hardware donation of a Tesla K40 GPU unit, Turkey.en_US
dc.description.abstractMany deep neural networks are built by using stacked convolutional layers of fixed and single size (often 3 × 3) kernels. This paper describes a method for learning the size of convolutional kernels to provide varying size kernels in a single layer. The method utilizes a differentiable, and therefore backpropagation-trainable Gaussian envelope which can grow or shrink in a base grid. Our experiments compared the proposed adaptive layers to ordinary convolution layers in a simple two-layer network, a deeper residual network, and a U-Net architecture. The results in the popular image classification datasets such as MNIST, MNIST-CLUTTERED, CIFAR-10, Fashion, and ‘‘Faces in the Wild’’ showed that the adaptive kernels can provide statistically significant improvements on ordinary convolution kernels. A segmentation experiment in the Oxford-Pets dataset demonstrated that replacing ordinary convolution layers in a U-shaped network with 7 × 7 adaptive layers can improve its learning performance and ability to generalize.en_US
dc.description.sponsorshipScientific and Technological Research Council of Turkey programmeen_US
dc.description.sponsorshipIsik Universityen_US
dc.description.sponsorshipNVIDIAen_US
dc.description.versionPublisher's Versionen_US
dc.identifier.citationTek, F. B., Çam, İ. & Karlı, D. (2021). Adaptive convolution kernel for artificial neural networks. Journal of Visual Communication and Image Representation, 75, 1-11.doi:10.1016/j.jvcir.2020.103015en_US
dc.identifier.doi10.1016/j.jvcir.2020.103015
dc.identifier.endpage11
dc.identifier.issn1047-3203
dc.identifier.issn1095-9076
dc.identifier.scopus2-s2.0-85099235848
dc.identifier.scopusqualityQ1
dc.identifier.startpage1
dc.identifier.urihttps://hdl.handle.net/11729/3075
dc.identifier.urihttp://dx.doi.org/10.1016/j.jvcir.2020.103015
dc.identifier.volume75
dc.identifier.wosWOS:000633494400002
dc.identifier.wosqualityQ2
dc.indekslendigikaynakWeb of Scienceen_US
dc.indekslendigikaynakScopusen_US
dc.indekslendigikaynakScience Citation Index Expanded (SCI-EXPANDED)en_US
dc.institutionauthorTek, Faik Borayen_US
dc.institutionauthorÇam, İlkeren_US
dc.institutionauthorKarlı, Denizen_US
dc.institutionauthorid0000-0002-8649-6013
dc.institutionauthorid0000-0002-5639-0648
dc.language.isoenen_US
dc.peerreviewedYesen_US
dc.publicationstatusPublisheden_US
dc.publisherAcademic Press Inc.en_US
dc.relation.ispartofJournal of Visual Communication and Image Representationen_US
dc.relation.publicationcategoryMakale - Uluslararası Hakemli Dergi - Kurum Öğretim Elemanıen_US
dc.rightsinfo:eu-repo/semantics/closedAccessen_US
dc.subjectAdaptive convolutionen_US
dc.subjectImage classificationen_US
dc.subjectMulti-scale convolutionen_US
dc.subjectResidual networksen_US
dc.subjectBackpropagationen_US
dc.subjectClassification (of information)en_US
dc.subjectConvolutionen_US
dc.subjectDeep learningen_US
dc.subjectDeep neural networksen_US
dc.subjectImage enhancementen_US
dc.subjectLearning systemsen_US
dc.subjectNetwork layersen_US
dc.subjectAdaptive kernelsen_US
dc.subjectClassification datasetsen_US
dc.subjectConvolution kernelen_US
dc.subjectConvolutional kernelen_US
dc.subjectGaussian envelopeen_US
dc.subjectLearning performanceen_US
dc.subjectNET architectureen_US
dc.subjectTwo-layer networken_US
dc.subjectMultilayer neural networksen_US
dc.titleAdaptive convolution kernel for artificial neural networksen_US
dc.typeArticleen_US

Dosyalar

Orijinal paket
Listeleniyor 1 - 1 / 1
Küçük Resim Yok
İsim:
3075.pdf
Boyut:
3.93 MB
Biçim:
Adobe Portable Document Format
Açıklama:
Publisher's Version
İndir
Lisans paketi
Listeleniyor 1 - 1 / 1
Küçük Resim Yok
İsim:
license.txt
Boyut:
1.44 KB
Biçim:
Item-specific license agreed upon to submission
Açıklama:
İndir

Koleksiyon

MF - Makale Koleksiyonu | Bilgisayar Mühendisliği Bölümü / Department of Computer Engineering
FEF - Makale Koleksiyonu | Matematik Bölümü / Department of Mathematics
Scopus İndeksli Yayınlar Koleksiyonu
WoS İndeksli Yayınlar Koleksiyonu