Using digital masks to enhance the bandwidth tolerance and improve利用数字掩模提高带宽容差和改善 片上储层计算系统的性能

Using digital masks to enhance the bandwidth tolerance and improve利用数字掩模提高带宽容差和改善 片上储层计算系统的性能

ID:39076718

大小:694.73 KB

页数:7页

时间:2019-06-24

Using digital masks to enhance the bandwidth tolerance and improve利用数字掩模提高带宽容差和改善 片上储层计算系统的性能_第1页
Using digital masks to enhance the bandwidth tolerance and improve利用数字掩模提高带宽容差和改善 片上储层计算系统的性能_第2页
Using digital masks to enhance the bandwidth tolerance and improve利用数字掩模提高带宽容差和改善 片上储层计算系统的性能_第3页
Using digital masks to enhance the bandwidth tolerance and improve利用数字掩模提高带宽容差和改善 片上储层计算系统的性能_第4页
Using digital masks to enhance the bandwidth tolerance and improve利用数字掩模提高带宽容差和改善 片上储层计算系统的性能_第5页
资源描述:

《Using digital masks to enhance the bandwidth tolerance and improve利用数字掩模提高带宽容差和改善 片上储层计算系统的性能》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库

1、>REPLACETHISLINEWITHYOURPAPERIDENTIFICATIONNUMBER(DOUBLE-CLICKHERETOEDIT)<1Usingdigitalmaskstoenhancethebandwidthtoleranceandimproveperformanceinon-chipreservoircomputingsystemsBendixSchneider,JoniDambre,Member,IEEEandPeterBienstman,Member,IEEEonconventionalcomputers.Aspecialroleisattributedtoa

2、ll-Abstract—ReservoirComputing(RC)isacomputingschemeopticalRCset-upsduetotheirinherenthighbandwidthwhichrelatedtorecurrentneuralnetworktheory.Asamodelforallowsforatremendousincreaseincomputingspeed.Aneuralactivityinthebrainitattractsalotofattention,especiallynetworkofintegratedSOAs(semiconductor

3、opticalbecauseofitsverysimpletrainingmethod.However,buildingaamplifiers)[14],[15],asinglenon-linearnodeinadelayedfunctional,on-chip,photonicimplementationofRCremainsafeedbackloop[16],andasemiconductorlaserdiodesubjecttochallenge.ScalingdelaylinesdownfromopticalfibrescaletochipscaleresultsinRCsys

4、temsthatcomputefaster,butatthedelayedself-feedback[17]havebeenproposedamongothers.sametimerequirethattheinputsignalsarealsoscaledupinTheon-chipphotonicintegrationofareservoiroffersthespeed,whichmightbeimpracticalorexpensive.Inthispaper,weadvantageofminiaturisation,cost-effectivenessandtheshowtha

5、tthisproblemcanbealleviatedbyamaskedRCsystempossibilityformassfabrication.However,itcanonlyperforminwhichtheamplitudeoftheinputsignalismodulatedbyausefulcomputationifitsinternaltimescalesmatchthebinary-valuedmask.Foraspeechrecognitiontaskwetimescaleoftheinputsignalathand.Itisdesirabletopreserved

6、emonstratethatthenecessaryinputsampleratecanbeafactorof40smallerthaninaconventionalRCsystem.Additionally,wetheinherentlyhighbandwidththatisofferedbythephotonicalsoshowthatlineardiscriminantanalysisandinputmatriximplementation.Yet,thischoicecompelstimescaleoptimisationisawell-performingalternativ

7、etolinearregressioncompressionformanyreal-timeapplications,e.g.speechforreservoirtraining.recognition,andimpliestheuseofhigh-speed,analoguewaveformgeneratorsasprerequisitesforRC.GiventhattheseIndexTerms—Photonicreservoircomp

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。