Apache MXNet 2.0：连接深度学习与传统机器学习.pdf

1、GTCApache MXNet 2.0Bridging Deep Learning and Machine LearningSheng ZhaSr.Applied Scientist Amazon AImxnetBUZSCzhashengapache.orgINCUBATOR#page#GTCAgendaCommunitys AI Framework fragmentation Apache MXNet 2.0 previewEcosystemmxnet#page#GTCApache MXNetCommunity 942 contributors19.1k stars，6.8k forks6.

2、1k commits since joining Apache Incubatormxnet#page#ALREADY TRUSTED BYmediaNET(intenVIDIA.adocomoProneeracer700elecomKASTCarnegi谢GeorgianSMalleMellongumgum9dely SIGOPTusimplePartnersUniversityABOREALISAIbanjoViSENZE美 pixmCiaocimpressEMOY中站MOQIYEOREOuE轻3EEVAclusteroneLOHIKATWOSENSE.AICmeWOLFRAMNaZIFF

3、BASLER7cometMAGNETICURALATEEAGLEVEWSo海wiisO EUROPACEALGORITHMIAllByteDanceFinn AIScling#page#GTCGet involvedHow to get started in MXNet？ Social media Stay connectedapachemxnetTwitter devmxnet.apache.orgXMMediumapache-mxnetthe- #mxnetYoutubeapachemxnet口19.opengoodfirstissueissuesLinkedIn apache-mxnet

4、Roadmap白Bilibili ApacheMXNet Improved C+ and Java bindingshttps:/mxnet.apache.org/versions/master/community TVM Relay integrationmxnet#page#GTCAl Framework Fragmentationmxnet#page#GTCDiverse，yet fragmentedH0BOONN0teAIOpenSource Landscape.Source:https:/landscape.lfai.foundation/mxnet#page#GTC“Fragmen

5、tation”lack of interoperabilitylack of common interface designmxnet#page#GTCWhy is fragmentation costly2o Userss Framework developerss Third-party developersmxnet#page#GTCWhy is fragmentation costly？Learning a SkillUsersTime Invested争Ts Framework developerss Third-party developers18Level of SkillCod

6、e baseFragmentation creates lock-inmxnet#page#GTCWhy is fragmentation costly？s Users Framework developersMinimums Thirdparty developersSource:http/ forces attentionto the shortest plankmxnet#page#GTCWhy is fragmentation costly2s Userss Framework developers Third-party developersFragmentation creates