A memristive all-inclusive hypernetwork for parallel analog deployment of full search space architectures

<p dir="ltr">In recent years, there has been a significant advancement in memristor-based neural networks, positioning them as a pivotal processing-in-memory deployment architecture for a wide array of deep learning applications. Within this realm of progress, the emerging parallel a...

وصف كامل

محفوظ في:
التفاصيل البيبلوغرافية
المؤلف الرئيسي: Bo Lyu (16522643) (author)
مؤلفون آخرون: Yin Yang (35103) (author), Yuting Cao (4231810) (author), Tuo Shi (3137919) (author), Yiran Chen (4787181) (author), Tingwen Huang (7168691) (author), Shiping Wen (7168688) (author)
منشور في: 2024
الموضوعات:
الوسوم: إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
_version_ 1864513509702762496
author Bo Lyu (16522643)
author2 Yin Yang (35103)
Yuting Cao (4231810)
Tuo Shi (3137919)
Yiran Chen (4787181)
Tingwen Huang (7168691)
Shiping Wen (7168688)
author2_role author
author
author
author
author
author
author_facet Bo Lyu (16522643)
Yin Yang (35103)
Yuting Cao (4231810)
Tuo Shi (3137919)
Yiran Chen (4787181)
Tingwen Huang (7168691)
Shiping Wen (7168688)
author_role author
dc.creator.none.fl_str_mv Bo Lyu (16522643)
Yin Yang (35103)
Yuting Cao (4231810)
Tuo Shi (3137919)
Yiran Chen (4787181)
Tingwen Huang (7168691)
Shiping Wen (7168688)
dc.date.none.fl_str_mv 2024-04-19T12:00:00Z
dc.identifier.none.fl_str_mv 10.1016/j.neunet.2024.106312
dc.relation.none.fl_str_mv https://figshare.com/articles/journal_contribution/A_memristive_all-inclusive_hypernetwork_for_parallel_analog_deployment_of_full_search_space_architectures/26491063
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
Distributed computing and systems software
Machine learning
Memristor
Convolutional neural network
Neural architecture search
Hypernetwork
dc.title.none.fl_str_mv A memristive all-inclusive hypernetwork for parallel analog deployment of full search space architectures
dc.type.none.fl_str_mv Text
Journal contribution
info:eu-repo/semantics/publishedVersion
text
contribution to journal
description <p dir="ltr">In recent years, there has been a significant advancement in memristor-based neural networks, positioning them as a pivotal processing-in-memory deployment architecture for a wide array of deep learning applications. Within this realm of progress, the emerging parallel analog memristive platforms are prominent for their ability to generate multiple feature maps in a single processing cycle. However, a notable limitation is that they are specifically tailored for neural networks with fixed structures. As an orthogonal direction, recent research reveals that neural architecture should be specialized for tasks and deployment platforms. Building upon this, the neural architecture search (NAS) methods effectively explore promising architectures in a large design space. However, these NAS-based architectures are generally heterogeneous and diversified, making it challenging for deployment on current single-prototype, customized, parallel analog memristive hardware circuits. Therefore, investigating memristive analog deployment that overrides the full search space is a promising and challenging problem. Inspired by this, and beginning with the <i>DARTS</i> search space, we study the memristive hardware design of primitive operations and propose the memristive all-inclusive hypernetwork that covers 2 × 1 0 <sup>25</sup> network architectures. Our computational simulation results on 3 representative architectures (<i>DARTS-V1, DARTS-V2, PDARTS</i>) show that our memristive all-inclusive hypernetwork achieves promising results on the CIFAR10 dataset (89.2 % of <i>PDARTS</i> with 8-bit quantization precision), and is compatible with all architectures in the <i>DARTS </i>full-space. The hardware performance simulation indicates that the memristive all-inclusive hypernetwork costs slightly more resource consumption (nearly the same in power, 22 % ∼ 25 % increase in Latency, 1 . 5 × in Area) relative to the individual deployment, which is reasonable and may reach a tolerable trade-off deployment scheme for industrial scenarios.</p><h2>Other Information</h2><p dir="ltr">Published in: Neural Networks<br>License: <a href="http://creativecommons.org/licenses/by/4.0/" target="_blank">http://creativecommons.org/licenses/by/4.0/</a><br>See article on publisher's website: <a href="https://dx.doi.org/10.1016/j.neunet.2024.106312" target="_blank">https://dx.doi.org/10.1016/j.neunet.2024.106312</a></p>
eu_rights_str_mv openAccess
id Manara2_7488442e34c07e4351ba993d3320cbbb
identifier_str_mv 10.1016/j.neunet.2024.106312
network_acronym_str Manara2
network_name_str Manara2
oai_identifier_str oai:figshare.com:article/26491063
publishDate 2024
repository.mail.fl_str_mv
repository.name.fl_str_mv
repository_id_str
rights_invalid_str_mv CC BY 4.0
spelling A memristive all-inclusive hypernetwork for parallel analog deployment of full search space architecturesBo Lyu (16522643)Yin Yang (35103)Yuting Cao (4231810)Tuo Shi (3137919)Yiran Chen (4787181)Tingwen Huang (7168691)Shiping Wen (7168688)Information and computing sciencesDistributed computing and systems softwareMachine learningMemristorConvolutional neural networkNeural architecture searchHypernetwork<p dir="ltr">In recent years, there has been a significant advancement in memristor-based neural networks, positioning them as a pivotal processing-in-memory deployment architecture for a wide array of deep learning applications. Within this realm of progress, the emerging parallel analog memristive platforms are prominent for their ability to generate multiple feature maps in a single processing cycle. However, a notable limitation is that they are specifically tailored for neural networks with fixed structures. As an orthogonal direction, recent research reveals that neural architecture should be specialized for tasks and deployment platforms. Building upon this, the neural architecture search (NAS) methods effectively explore promising architectures in a large design space. However, these NAS-based architectures are generally heterogeneous and diversified, making it challenging for deployment on current single-prototype, customized, parallel analog memristive hardware circuits. Therefore, investigating memristive analog deployment that overrides the full search space is a promising and challenging problem. Inspired by this, and beginning with the <i>DARTS</i> search space, we study the memristive hardware design of primitive operations and propose the memristive all-inclusive hypernetwork that covers 2 × 1 0 <sup>25</sup> network architectures. Our computational simulation results on 3 representative architectures (<i>DARTS-V1, DARTS-V2, PDARTS</i>) show that our memristive all-inclusive hypernetwork achieves promising results on the CIFAR10 dataset (89.2 % of <i>PDARTS</i> with 8-bit quantization precision), and is compatible with all architectures in the <i>DARTS </i>full-space. The hardware performance simulation indicates that the memristive all-inclusive hypernetwork costs slightly more resource consumption (nearly the same in power, 22 % ∼ 25 % increase in Latency, 1 . 5 × in Area) relative to the individual deployment, which is reasonable and may reach a tolerable trade-off deployment scheme for industrial scenarios.</p><h2>Other Information</h2><p dir="ltr">Published in: Neural Networks<br>License: <a href="http://creativecommons.org/licenses/by/4.0/" target="_blank">http://creativecommons.org/licenses/by/4.0/</a><br>See article on publisher's website: <a href="https://dx.doi.org/10.1016/j.neunet.2024.106312" target="_blank">https://dx.doi.org/10.1016/j.neunet.2024.106312</a></p>2024-04-19T12:00:00ZTextJournal contributioninfo:eu-repo/semantics/publishedVersiontextcontribution to journal10.1016/j.neunet.2024.106312https://figshare.com/articles/journal_contribution/A_memristive_all-inclusive_hypernetwork_for_parallel_analog_deployment_of_full_search_space_architectures/26491063CC BY 4.0info:eu-repo/semantics/openAccessoai:figshare.com:article/264910632024-04-19T12:00:00Z
spellingShingle A memristive all-inclusive hypernetwork for parallel analog deployment of full search space architectures
Bo Lyu (16522643)
Information and computing sciences
Distributed computing and systems software
Machine learning
Memristor
Convolutional neural network
Neural architecture search
Hypernetwork
status_str publishedVersion
title A memristive all-inclusive hypernetwork for parallel analog deployment of full search space architectures
title_full A memristive all-inclusive hypernetwork for parallel analog deployment of full search space architectures
title_fullStr A memristive all-inclusive hypernetwork for parallel analog deployment of full search space architectures
title_full_unstemmed A memristive all-inclusive hypernetwork for parallel analog deployment of full search space architectures
title_short A memristive all-inclusive hypernetwork for parallel analog deployment of full search space architectures
title_sort A memristive all-inclusive hypernetwork for parallel analog deployment of full search space architectures
topic Information and computing sciences
Distributed computing and systems software
Machine learning
Memristor
Convolutional neural network
Neural architecture search
Hypernetwork