Basit öğe kaydını göster

dc.contributor.authorKaytan, Mustafa
dc.contributor.authorAydilek, İbrahim Berkan
dc.contributor.authorYeroğlu, Celaleddin
dc.date.accessioned2024-01-15T12:11:38Z
dc.date.available2024-01-15T12:11:38Z
dc.date.issued2023en_US
dc.identifier.citationKaytan, M., Aydilek, İ.B., Yeroğlu, C. (2023). Gish: a novel activation function for image classification. Neural Computing and Applications, 35 (34), pp. 24259-24281. https://doi.org/10.1007/s00521-023-09035-5en_US
dc.identifier.issn0941-0643
dc.identifier.issn1433-3058
dc.identifier.urihttps://doi.org/10.1007/s00521-023-09035-5
dc.identifier.urihttps://hdl.handle.net/20.500.12508/3005
dc.description.abstractIn Convolutional Neural Networks (CNNs), the selection and use of appropriate activation functions is of critical importance. It has been seen that the Rectified Linear Unit (ReLU) is widely used in many CNN models. Looking at the recent studies, it has been seen that some non-monotonic activation functions are gradually moving towards becoming the new standard to improve the performance of CNN models. It has been observed that some non-monotonic activation functions such as Swish, Mish, Logish and Smish are used to obtain successful results in various deep learning models. However, only a few of them have been widely used in most of the studies. Inspired by them, in this study, a new activation function named Gish, whose mathematical model can be represented by y=x·ln(2-e-ex) , which can overcome other activation functions with its good properties, is proposed. The variable x is used to contribute to a strong regulation effect of negative output. The logarithm operation is done to reduce the numerical range of the expression (2-e-ex) . To present our contributions in this work, various experiments were conducted on different network models and datasets to evaluate the performance of Gish. With the experimental results, 98.7% success was achieved with the EfficientNetB4 model in the MNIST dataset, 86.5% with the EfficientNetB5 model in the CIFAR-10 dataset and 90.8% with the EfficientNetB6 model in the SVHN dataset. The obtained performances were shown to be higher than Swish, Mish, Logish and Smish. These results confirm the effectiveness and performance of Gish.en_US
dc.language.isoengen_US
dc.publisherSpringeren_US
dc.relation.isversionof10.1007/s00521-023-09035-5en_US
dc.rightsinfo:eu-repo/semantics/closedAccessen_US
dc.subjectConvolutional neural networken_US
dc.subjectGishen_US
dc.subjectImage classificationen_US
dc.subjectNonmonotonic activation functionen_US
dc.subject.classificationObject Detection
dc.subject.classificationDeep Learning
dc.subject.classificationIOU
dc.subject.classificationElectrical Engineering, Electronics & Computer Science - Computer Vision & Graphics - Genome Rearrangement
dc.subject.otherConvergence
dc.subject.otherNetworks
dc.subject.otherNeurons
dc.subject.otherSpeed
dc.subject.otherChemical activation
dc.subject.otherConvolution
dc.subject.otherConvolutional neural networks
dc.subject.otherDeep learning
dc.subject.otherNeural network models
dc.subject.otherActivation functions
dc.subject.otherConvolutional neural network
dc.subject.otherGish
dc.subject.otherImages classification
dc.subject.otherLinear units
dc.subject.otherMonotonics
dc.subject.otherNeural network model
dc.subject.otherNonmonotonic
dc.subject.otherNonmonotonic activation function
dc.subject.otherPerformance
dc.subject.otherImage classification
dc.titleGish: a novel activation function for image classificationen_US
dc.typearticleen_US
dc.relation.journalNeural Computing and Applicationsen_US
dc.contributor.departmentMühendislik ve Doğa Bilimleri Fakültesi -- Bilgisayar Mühendisliği Bölümüen_US
dc.identifier.volume35en_US
dc.identifier.issue34en_US
dc.identifier.startpage24259en_US
dc.identifier.endpage24281en_US
dc.relation.publicationcategoryMakale - Uluslararası Hakemli Dergi - Kurum Öğretim Elemanıen_US
dc.contributor.isteauthorYeroğlu, Celaleddin
dc.relation.indexWeb of Science - Scopusen_US
dc.relation.indexWeb of Science Core Collection - Science Citation Index Expanded


Bu öğenin dosyaları:

Thumbnail

Bu öğe aşağıdaki koleksiyon(lar)da görünmektedir.

Basit öğe kaydını göster