FatNet:使用全卷积光学神经网络进行分类的高分辨率核

IF 3.1 Q2 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE AI (Basel, Switzerland) Pub Date : 2023-04-03 DOI:10.3390/ai4020018
Riad Ibadulla, Thomas M. Chen, Constantino Carlos Reyes-Aldasoro
{"title":"FatNet:使用全卷积光学神经网络进行分类的高分辨率核","authors":"Riad Ibadulla, Thomas M. Chen, Constantino Carlos Reyes-Aldasoro","doi":"10.3390/ai4020018","DOIUrl":null,"url":null,"abstract":"This paper describes the transformation of a traditional in silico classification network into an optical fully convolutional neural network with high-resolution feature maps and kernels. When using the free-space 4f system to accelerate the inference speed of neural networks, higher resolutions of feature maps and kernels can be used without the loss in frame rate. We present FatNet for the classification of images, which is more compatible with free-space acceleration than standard convolutional classifiers. It neglects the standard combination of convolutional feature extraction and classifier dense layers by performing both in one fully convolutional network. This approach takes full advantage of the parallelism in the 4f free-space system and performs fewer conversions between electronics and optics by reducing the number of channels and increasing the resolution, making this network faster in optics than off-the-shelf networks. To demonstrate the capabilities of FatNet, it was trained with the CIFAR100 dataset on GPU and the simulator of the 4f system. A comparison of the results against ResNet-18 shows 8.2 times fewer convolution operations at the cost of only 6% lower accuracy. This demonstrates that the optical implementation of FatNet results in significantly faster inference than the optical implementation of the original ResNet-18. These are promising results for the approach of training deep learning with high-resolution kernels in the direction toward the upcoming optics era.","PeriodicalId":93633,"journal":{"name":"AI (Basel, Switzerland)","volume":null,"pages":null},"PeriodicalIF":3.1000,"publicationDate":"2023-04-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":"{\"title\":\"FatNet: High-Resolution Kernels for Classification Using Fully Convolutional Optical Neural Networks\",\"authors\":\"Riad Ibadulla, Thomas M. Chen, Constantino Carlos Reyes-Aldasoro\",\"doi\":\"10.3390/ai4020018\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"This paper describes the transformation of a traditional in silico classification network into an optical fully convolutional neural network with high-resolution feature maps and kernels. When using the free-space 4f system to accelerate the inference speed of neural networks, higher resolutions of feature maps and kernels can be used without the loss in frame rate. We present FatNet for the classification of images, which is more compatible with free-space acceleration than standard convolutional classifiers. It neglects the standard combination of convolutional feature extraction and classifier dense layers by performing both in one fully convolutional network. This approach takes full advantage of the parallelism in the 4f free-space system and performs fewer conversions between electronics and optics by reducing the number of channels and increasing the resolution, making this network faster in optics than off-the-shelf networks. To demonstrate the capabilities of FatNet, it was trained with the CIFAR100 dataset on GPU and the simulator of the 4f system. A comparison of the results against ResNet-18 shows 8.2 times fewer convolution operations at the cost of only 6% lower accuracy. This demonstrates that the optical implementation of FatNet results in significantly faster inference than the optical implementation of the original ResNet-18. These are promising results for the approach of training deep learning with high-resolution kernels in the direction toward the upcoming optics era.\",\"PeriodicalId\":93633,\"journal\":{\"name\":\"AI (Basel, Switzerland)\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":3.1000,\"publicationDate\":\"2023-04-03\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"1\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"AI (Basel, Switzerland)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.3390/ai4020018\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"AI (Basel, Switzerland)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.3390/ai4020018","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
引用次数: 1

摘要

本文描述了将传统的计算机分类网络转化为具有高分辨率特征映射和核的光学全卷积神经网络。当使用自由空间4f系统加速神经网络的推理速度时,可以在不损失帧率的情况下使用更高分辨率的特征映射和核。我们提出了用于图像分类的FatNet,它比标准卷积分类器更兼容自由空间加速。它忽略了卷积特征提取和分类器密集层的标准组合,在一个全卷积网络中执行。这种方法充分利用了4f自由空间系统的并行性,并通过减少通道数量和提高分辨率来减少电子和光学之间的转换,使该网络在光学上比现成的网络更快。为了证明FatNet的能力,在GPU上使用CIFAR100数据集和4f系统的模拟器对其进行了训练。与ResNet-18的结果比较显示,卷积操作减少了8.2倍,而准确率仅降低了6%。这表明FatNet的光学实现比原始ResNet-18的光学实现的推理速度要快得多。这些都是在即将到来的光学时代,用高分辨率核训练深度学习方法的有希望的结果。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
查看原文
分享 分享
微信好友 朋友圈 QQ好友 复制链接
本刊更多论文
FatNet: High-Resolution Kernels for Classification Using Fully Convolutional Optical Neural Networks
This paper describes the transformation of a traditional in silico classification network into an optical fully convolutional neural network with high-resolution feature maps and kernels. When using the free-space 4f system to accelerate the inference speed of neural networks, higher resolutions of feature maps and kernels can be used without the loss in frame rate. We present FatNet for the classification of images, which is more compatible with free-space acceleration than standard convolutional classifiers. It neglects the standard combination of convolutional feature extraction and classifier dense layers by performing both in one fully convolutional network. This approach takes full advantage of the parallelism in the 4f free-space system and performs fewer conversions between electronics and optics by reducing the number of channels and increasing the resolution, making this network faster in optics than off-the-shelf networks. To demonstrate the capabilities of FatNet, it was trained with the CIFAR100 dataset on GPU and the simulator of the 4f system. A comparison of the results against ResNet-18 shows 8.2 times fewer convolution operations at the cost of only 6% lower accuracy. This demonstrates that the optical implementation of FatNet results in significantly faster inference than the optical implementation of the original ResNet-18. These are promising results for the approach of training deep learning with high-resolution kernels in the direction toward the upcoming optics era.
求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
CiteScore
7.20
自引率
0.00%
发文量
0
审稿时长
11 weeks
期刊最新文献
Can Artificial Intelligence Aid Diagnosis by Teleguided Point-of-Care Ultrasound? A Pilot Study for Evaluating a Novel Computer Algorithm for COVID-19 Diagnosis Using Lung Ultrasound. Chatbots Put to the Test in Math and Logic Problems: A Comparison and Assessment of ChatGPT-3.5, ChatGPT-4, and Google Bard Deep Learning Performance Characterization on GPUs for Various Quantization Frameworks From Trustworthy Principles to a Trustworthy Development Process: The Need and Elements of Trusted Development of AI Systems Algorithms for All: Can AI in the Mortgage Market Expand Access to Homeownership?
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
现在去查看 取消
×
提示
确定
0
微信
客服QQ
Book学术公众号 扫码关注我们
反馈
×
意见反馈
请填写您的意见或建议
请填写您的手机或邮箱
已复制链接
已复制链接
快去分享给好友吧!
我知道了
×
扫码分享
扫码分享
Book学术官方微信
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术
文献互助 智能选刊 最新文献 互助须知 联系我们:info@booksci.cn
Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。
Copyright © 2023 Book学术 All rights reserved.
ghs 京公网安备 11010802042870号 京ICP备2023020795号-1