google从永慧航空航天模式识别与智能系统_第1页
google从永慧航空航天模式识别与智能系统_第2页
google从永慧航空航天模式识别与智能系统_第3页
google从永慧航空航天模式识别与智能系统_第4页
google从永慧航空航天模式识别与智能系统_第5页
全文预览已结束

下载本文档

kok电子竞技权说明：本文档由用户提供并上传，收益归属内容提供方，若内容存在侵权，请进行举报或认领

kok电子竞技:文档简介

准确率达80%以上。基于深度学习的手写识别研究三星通信研究院实习 模型实现词汇聚类和语言模型的建立，并进行优化。英语准确率达90%以上，欧洲语言准确率平均达YonghuiCong,ZengchangQin,JingYuandTaoWan.Cross-ModalInformationRetrievalACaseStudyonChineseWikipedia.InInternationalConferenceonAdvancedDataMiningandApplications2012:15-26.EIJingYu,YonghuiCongZengchangQinandTaoWan.Cross-ModalTopicCorrelationsforMultimediaRetrieval.InInternationalConferenceonPatternRecognition,2012:246-249.(EI检索，已发表),TopicCorrelationModelforCross-ModalMultimediaInformationRetrievalPatternAnalysisandApplications（SCI检索，在审，第三作者）（NMOE院校级奖项A类，北航专业二等奖学Yonghui TEL: 2012.9-BeihangUniversity,Master,MachineLearning,Intelligentcomputingandmachinelearning2008.9-Familiarwithalgorithmsaboutcomputervision,andimageprocessing,includingcontent-basedimageretrieval,cross-modalinformationretrievalandimagefeatureextraction.Ihavealgorithmdevelopmentexperiencesondesigningpracticalproject.Familiarwithnaturallanguageprocessingalgorithmssuchastextmodeling,semanticunderstandinganalysis,syntacticstructure,topicmodelandlanguagemodel.Familiarwithbasicalgorithmsondataminingandmachinelearning,includingnaiveBayesclassifier,SVM,Adaboost,PCA,K-meansetc.Ihavemathematicsfoundationsonstatistics,probabilitytheory,matrixtheoryandsoon.FamiliarwithC/C++,Matlab,Python,OpenCVandLinuxdevelopmentenvironment.,understandingPerl,ShellandVisiblecomponentsDetectionofUrineSedimentInternshipinSiemensCTDescription:visiblecomponentsDetectionofUrineSediment.Mainlyresponsibleforresearchandanalysisonfeatureextractionandclassificationalgorithmsforcastcells.SIFT,LBP,HOGfeaturesandSVM,Adaboostclassifiersareusedintheproject.Ihavecompletedavalidclassificationoncastcells,theaccuracyrategetsupto80%.Matlab,PythonandOpenCVaremainlyused.HandwritingRecognitionBeijingSamsungTechnology Description:HandwritingRecognition.MainlyresponsiblefortrainingandoptimizationoflanguageUsingthen-gramandn-classalgorithmforwordclusteringandtraininglanguagemodel.ImprovingtheaccuracyofhandwritingrecognitionbymodifyingtheclusteringalgorithmandcombiningHTKandacousticmodel.OptimizationthelanguagemodelwithHLRescorewhichimprovestheresultby20%.C++,pythonaremainlyused.Cross-modalinformationretrievalbasedonprobability Description:Implementingcross-modalinformationretrievalviaaninputmediaform(suchasimages),toretrieveotherformsofmedia(suchastext)whichdescribesthesamecontent.IwasprimarilyresponsiblefortheexperimentsontheEnglishdatabase.IestablishedaChinesedatabasewithimagesandtexts,extractedimagefeaturesandtrainedtopicmodelbasedonwordsandcharactersrespectively.SIFTfeature,LDAandSVMaremainlyusedintheproject.Wehavebuildanovelprobabilitymodelbasedonclassinformation.ThisisthefirstChinesecaserealizingcross-modalinformationretrieval,andgettingabetterretrievalperformance.ThemainprogramlanguagesusedareMatlabandPython.TopicModelingonChinese 2013Description:AsChinesedifferentfromotherwesternlanguages,weanalyzethesyntacticandsemanticsstructureoftheChineselanguage.WeanalyzetopicdistributionbasedonChinesewordsandcharactersasthebasicunitandestablishthemappingrelationshipbetweencharactersandwords,inordertoestablishamoreeffectivesemanticmodel.Theworkisprocessedasthreeperiods:buildingtherelationshipbetweenwordsandcharacters;placingassymetricpriorbeforethedocument-topicdistributionandtopic-worddistribution;improvingthetopicmodelbasedonn-gramandclassinformation.MainlyusingLDAandGibbsSamplingandn-gramalgorithm.Wehaveverifiedthatthecharacter-basedLDAcanbemoreeffectivethantheword-basedLDA,andestablishedanimprovedtopicmodelbasedontheChinesecharacter-wordrelationshipandn-gram.Theeffectivenessofmodelhavebeentestified.ThemainlanguageisPythonandMatlab.YonghuiCong,ZengchangQin,JingYuandTaoWan.Cross-ModalInformationRetrieval-ACaseStudyonChineseWikipedia.InternationalConferenceonAdvancedDataMiningandApplications,2012:15-26.(EI)JingYu,YonghuiCong,ZengchangQinandTaoWan.Cross-ModalTopicCorrelationsforMultimediaRetrieval.InternationalConferenceonPatternRecognition,2012:246-249.English

人人文库> 全部分类> 教育资料 > 课件下载

温馨提示

  • 1. 本站所有资源如无特殊说明，都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
  • 2. 本站的文档不包含任何第三方提供的附件图纸等，如果需要附件，请联系上传者。文件的所有权益归上传用户所有。
  • 3. 本站RAR压缩包中若带图纸，网页内容里面会有图纸预览，若没有图纸预览就没有图纸。
  • 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
  • 5. 人人文库网仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对用户上传分享的文档内容本身不做任何修改或编辑，并不能对任何下载内容负责。
  • 6. 下载文件中如有侵权或不适当内容，请与我们联系，我们立即纠正。
  • 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。

kok电子竞技:最新文档

评论

0/150

提交评论