Super Neurons

<p dir="ltr">Self-Organized Operational Neural Networks (Self-ONNs) have recently been proposed as new-generation neural network models with nonlinear learning units, i.e., the generative neurons that yield an elegant level of diversity; however, like its predecessor, conventional Co...

وصف كامل

محفوظ في:
التفاصيل البيبلوغرافية
المؤلف الرئيسي: Serkan Kiranyaz (3762058) (author)
مؤلفون آخرون: Junaid Malik (16869930) (author), Mehmet Yamac (17986993) (author), Mert Duman (17986996) (author), Ilke Adalioglu (17986999) (author), Esin Guldogan (17987002) (author), Turker Ince (14150610) (author), Moncef Gabbouj (2276533) (author)
منشور في: 2023
الموضوعات:
الوسوم: إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
_version_ 1864513527015800832
author Serkan Kiranyaz (3762058)
author2 Junaid Malik (16869930)
Mehmet Yamac (17986993)
Mert Duman (17986996)
Ilke Adalioglu (17986999)
Esin Guldogan (17987002)
Turker Ince (14150610)
Moncef Gabbouj (2276533)
author2_role author
author
author
author
author
author
author
author_facet Serkan Kiranyaz (3762058)
Junaid Malik (16869930)
Mehmet Yamac (17986993)
Mert Duman (17986996)
Ilke Adalioglu (17986999)
Esin Guldogan (17987002)
Turker Ince (14150610)
Moncef Gabbouj (2276533)
author_role author
dc.creator.none.fl_str_mv Serkan Kiranyaz (3762058)
Junaid Malik (16869930)
Mehmet Yamac (17986993)
Mert Duman (17986996)
Ilke Adalioglu (17986999)
Esin Guldogan (17987002)
Turker Ince (14150610)
Moncef Gabbouj (2276533)
dc.date.none.fl_str_mv 2023-10-09T12:00:00Z
dc.identifier.none.fl_str_mv 10.1109/tetci.2023.3314658
dc.relation.none.fl_str_mv https://figshare.com/articles/journal_contribution/Super_Neurons/25243378
dc.rights.none.fl_str_mv CC BY 4.0
info:eu-repo/semantics/openAccess
dc.subject.none.fl_str_mv Information and computing sciences
Artificial intelligence
Mathematical sciences
Numerical and computational mathematics
Kernel
Neurons
Training
Biological neural networks
Computational intelligence
Biological system modeling
Location awareness
Convolutional neural networks
generative neurons
non-localized kernels
operational neural networks
receptive field
dc.title.none.fl_str_mv Super Neurons
dc.type.none.fl_str_mv Text
Journal contribution
info:eu-repo/semantics/publishedVersion
text
contribution to journal
description <p dir="ltr">Self-Organized Operational Neural Networks (Self-ONNs) have recently been proposed as new-generation neural network models with nonlinear learning units, i.e., the generative neurons that yield an elegant level of diversity; however, like its predecessor, conventional Convolutional Neural Networks (CNNs), they still have a common drawback: localized (fixed) kernel operations. This severely limits the receptive field and information flow between layers and thus brings the necessity for deep and complex models. It is highly desired to improve the receptive field size without increasing the kernel dimensions. This requires a significant upgrade over the generative neurons to achieve the “non-localized kernel operations” for each connection between consecutive layers. In this article, we present superior (generative) neuron models (or super neurons in short) that allow random or learnable kernel shifts and thus can increase the receptive field size of each connection. The kernel localization process varies among the two super-neuron models. The first model assumes randomly localized kernels within a range and the second one learns (optimizes) the kernel locations during training. An extensive set of comparative evaluations against conventional and deformable convolutional, along with the generative neurons demonstrates that super neurons can empower Self-ONNs to achieve a superior learning and generalization capability with a minimal computational complexity burden. PyTorch implementation of Self-ONNs with super-neurons is now publically shared.</p><h2>Other Information</h2><p dir="ltr">Published in: IEEE Transactions on Emerging Topics in Computational Intelligence<br>License: <a href="https://creativecommons.org/licenses/by/4.0/" target="_blank">https://creativecommons.org/licenses/by/4.0/</a><br>See article on publisher's website: <a href="https://dx.doi.org/10.1109/tetci.2023.3314658" target="_blank">https://dx.doi.org/10.1109/tetci.2023.3314658</a></p>
eu_rights_str_mv openAccess
id Manara2_84292a0601d12f9b3de4f2e5650f4139
identifier_str_mv 10.1109/tetci.2023.3314658
network_acronym_str Manara2
network_name_str Manara2
oai_identifier_str oai:figshare.com:article/25243378
publishDate 2023
repository.mail.fl_str_mv
repository.name.fl_str_mv
repository_id_str
rights_invalid_str_mv CC BY 4.0
spelling Super NeuronsSerkan Kiranyaz (3762058)Junaid Malik (16869930)Mehmet Yamac (17986993)Mert Duman (17986996)Ilke Adalioglu (17986999)Esin Guldogan (17987002)Turker Ince (14150610)Moncef Gabbouj (2276533)Information and computing sciencesArtificial intelligenceMathematical sciencesNumerical and computational mathematicsKernelNeuronsTrainingBiological neural networksComputational intelligenceBiological system modelingLocation awarenessConvolutional neural networksgenerative neuronsnon-localized kernelsoperational neural networksreceptive field<p dir="ltr">Self-Organized Operational Neural Networks (Self-ONNs) have recently been proposed as new-generation neural network models with nonlinear learning units, i.e., the generative neurons that yield an elegant level of diversity; however, like its predecessor, conventional Convolutional Neural Networks (CNNs), they still have a common drawback: localized (fixed) kernel operations. This severely limits the receptive field and information flow between layers and thus brings the necessity for deep and complex models. It is highly desired to improve the receptive field size without increasing the kernel dimensions. This requires a significant upgrade over the generative neurons to achieve the “non-localized kernel operations” for each connection between consecutive layers. In this article, we present superior (generative) neuron models (or super neurons in short) that allow random or learnable kernel shifts and thus can increase the receptive field size of each connection. The kernel localization process varies among the two super-neuron models. The first model assumes randomly localized kernels within a range and the second one learns (optimizes) the kernel locations during training. An extensive set of comparative evaluations against conventional and deformable convolutional, along with the generative neurons demonstrates that super neurons can empower Self-ONNs to achieve a superior learning and generalization capability with a minimal computational complexity burden. PyTorch implementation of Self-ONNs with super-neurons is now publically shared.</p><h2>Other Information</h2><p dir="ltr">Published in: IEEE Transactions on Emerging Topics in Computational Intelligence<br>License: <a href="https://creativecommons.org/licenses/by/4.0/" target="_blank">https://creativecommons.org/licenses/by/4.0/</a><br>See article on publisher's website: <a href="https://dx.doi.org/10.1109/tetci.2023.3314658" target="_blank">https://dx.doi.org/10.1109/tetci.2023.3314658</a></p>2023-10-09T12:00:00ZTextJournal contributioninfo:eu-repo/semantics/publishedVersiontextcontribution to journal10.1109/tetci.2023.3314658https://figshare.com/articles/journal_contribution/Super_Neurons/25243378CC BY 4.0info:eu-repo/semantics/openAccessoai:figshare.com:article/252433782023-10-09T12:00:00Z
spellingShingle Super Neurons
Serkan Kiranyaz (3762058)
Information and computing sciences
Artificial intelligence
Mathematical sciences
Numerical and computational mathematics
Kernel
Neurons
Training
Biological neural networks
Computational intelligence
Biological system modeling
Location awareness
Convolutional neural networks
generative neurons
non-localized kernels
operational neural networks
receptive field
status_str publishedVersion
title Super Neurons
title_full Super Neurons
title_fullStr Super Neurons
title_full_unstemmed Super Neurons
title_short Super Neurons
title_sort Super Neurons
topic Information and computing sciences
Artificial intelligence
Mathematical sciences
Numerical and computational mathematics
Kernel
Neurons
Training
Biological neural networks
Computational intelligence
Biological system modeling
Location awareness
Convolutional neural networks
generative neurons
non-localized kernels
operational neural networks
receptive field