統(tǒng)一的大數(shù)據(jù)分析及AI應(yīng)用平臺(tái)_第1頁(yè)
統(tǒng)一的大數(shù)據(jù)分析及AI應(yīng)用平臺(tái)_第2頁(yè)
統(tǒng)一的大數(shù)據(jù)分析及AI應(yīng)用平臺(tái)_第3頁(yè)
統(tǒng)一的大數(shù)據(jù)分析及AI應(yīng)用平臺(tái)_第4頁(yè)
統(tǒng)一的大數(shù)據(jù)分析及AI應(yīng)用平臺(tái)_第5頁(yè)
已閱讀5頁(yè),還剩20頁(yè)未讀 繼續(xù)免費(fèi)閱讀

下載本文檔

版權(quán)說(shuō)明:本文檔由用戶提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權(quán),請(qǐng)進(jìn)行舉報(bào)或認(rèn)領(lǐng)

文檔簡(jiǎn)介

技術(shù)創(chuàng)新,變革未來(lái)統(tǒng)一的大數(shù)據(jù)分析及AI應(yīng)用平臺(tái)All

products,

computer

systems,

dates,

and

figures

are

preliminary

based

on

current

expectations,

and

are

subject

to

change

without

notice.2IntelGPUfutureAutomatedDrivingDedicatedMedia/VisionAcceleratio

DLnferencn I eDedicatedDLTrainingGraphics,Media&

Analytics2H’192H’19NNP-LNNP-IFlexible

IfneededDedicatededgeCloudDevIceOneSizeDoes

NotFitAll31Anopensourceversionisavailableat:01.org/openvinotoolkit*Othernamesandbrandsmaybeclaimedasthepropertyof

others.Developerpersonasshowaboverepresenttheprimaryuserbaseforeachrow,butarenot

mutually-exclusiveAllproducts,computersystems,dates,andfiguresarepreliminarybasedoncurrentexpectations,andaresubjecttochangewithout

notice.TOOLKITSAppdeveloperslibrariesDatascientistsKernelsLibrarydevelopersOpensourceplatformforbuildingE2EAnalytics&AIapplicationsonApacheSpark*withdistributedTensorFlow*,Keras*,

BigDLDeeplearninginferencedeploymentonCPU/GPU/FPGA/VPUforCaffe*,TensorFlow*,MXNet*,ONNX*,

Kaldi*Opensource,scalable,andextensibledistributeddeeplearningplatformbuiltonKubernetes

(BETA)Intel-optimized

FrameworksAndmoreframeworkoptimizationsunderwayincludingPaddlePaddle*,Chainer*,CNTK*&

othersPythonScikit-learnPandasNumPyRCartRandom

Foreste1071DistributedMlLib(on

Spark)MahoutIntel?

Distribution

for

Python*Inteldistributionoptimizedformachine

learningIntel?Data

Analytics

AccelerationLibrary

(DAAL)Highperformancemachinelearning&dataanalytics

libraryOpensourcecompilerfordeeplearningmodelcomputationsoptimizedformultipledevices(CPU,GPU,NNP)frommultipleframeworks(TF,MXNet,

ONNX)Intel?Math

Kernel

LibraryforDeep

NeuralNetworks

(MKL-DNN)OpensourceDNNfunctions

forCPU/integrated

graphicsMachine

learning Deep

learning*****SpeedUp

DevelopmentUsingOpenAI

SoftwareDistributed,

High-PerformanceDeepLearning

FrameworkforApache

Spark*/intel-analytics/bigdlAnalytics+AI

PlatformDistributedTensorFlow*,Keras*and

BigDLonApache

Spark*/intel-analytics/analytics-zooAI

onUnifyingAnalytics+AIonApache

Spark**Othernamesandbrandsmaybeclaimedasthepropertyof

others.WhyAnalytics

Zoo?Real-WorldML/DLApplicationsAreComplexDataAnalytics

Pipelines“Hidden

Technical

Debt

in

Machine

Learning

Systems”,Sculleyetal.,Google,NIPS2015

PaperLarge-ScaleImageRecognitionat

JD.com/en-us/articles/building-large-scale-image-feature-extraction-with-bigdl-at-jdcomChasmb/wDeepLearningandBigDataCommunitiesDeeplearning

expertsTheChasmReal-worldusers(bigdatausers,datascientists,analysts,

etc.)Distributed,

High-PerformanceDeepLearning

FrameworkforApache

Spark*/intel-analytics/bigdlAnalytics+AI

PlatformDistributedTensorFlow*,Keras*and

BigDLonApache

Spark*/intel-analytics/analytics-zooAI

onUnifyingAnalytics+AIonApache

Spark**Othernamesandbrandsmaybeclaimedasthepropertyof

others./en-us/videos/analytics-zoo-overviewAnalyticsZoo

VideoAnalyticsZoo:End-to-EndDLPipelineMadeEasyforBig

DataPrototypeonlaptopusingsample

dataExperimentonclusterswithhistory

dataDeploymentwithproduction,distribtued

bigdata

pipelines“Zero”codechangefromlaptoptodistributed

clusterDirectlyaccessingproductionbigdata

(Hadoop/Hive/HBase)Easilyprototypingtheend-to-end

pipelineSeamlesslydeployedonproductionbigdata

clustersWhatisAnalytics

Zoo?Analytics

ZooBERTtfpark:DistributedTF

onBigDatannframes:SparkDataframes&

MLPipelinesforDeep

LearningDistributedKerasw/autogradonBig

DataDistributedModelServing(batch,streaming&

online)Image

ClassificationObject

Detectionimage3D

imageTransformertextSeq2SeqUse

caseModelFeature

EngineeringHigh

LevelPipelinesBackend/Librarytime

seriesRecommendation Anomaly

Detection Text

Classification Text

Matching

End-to-End,

Integrated

Data

Analytics

+

AI

Platform /intel-analytics/analytics-zooKeras PyTorch BigDL NLP

Architect Apache

Spark Apache

FlinkMKLDNN OpenVINO Intel?Optane?

DCPMM DLBoost

(VNNI)TensorFlowRayAnalyticsZooUnifiedAnalytics+AIPlatformforBig

DataBuildend-to-enddeeplearningapplicationsforbig

dataDistributedTensorFlowon

SparkKerasAPI(withautograd&transferlearningsupport)on

Sparknnframes:nativeDLsupportforSparkDataFramesandML

PipelinesProductionizedeeplearningapplicationsforbigdataat

scalePlainJava/PythonmodelservingAPIs(w/OpenVINO

support)SupportWebServices,Spark,Flink,Storm,Kafka,etc.Out-of-the-box

solutionsBuilt-indeeplearningmodels,featureengineeringoperations,andreferenceusecasesDistributedTF&Kerason

SparkDatawranglingandanalysisusing

PySparkDeeplearning

modeldevelopmentusingTensorFlowor

KerasDistributedtraining

/inferenceon

Spark#pyspark

codetrain_rdd=spark.hadoopFile(…).map(…)dataset=

TFDataset.from_rdd(train_rdd,…)#tensorflow

codeimporttensorflowas

tfslim=

tf.contrib.slimimages,labels=

dataset.tensorswithslim.arg_scope(lenet.lenet_arg_scope()):logits,end_points=lenet.lenet(images,

…)loss=tf.reduce_mean(\tf.losses.sparse_softmax_cross_entropy(\logits=logits,

labels=labels))#distributedtrainingon

Sparkoptimizer=TFOptimizer.from_loss(loss,Adam(…))

\optimizer.optimize(end_trigger=MaxEpoch(5))WriteTensorFlowcodeinlineinPySpark

programSparkDataframe&MLPipelinefor

DL#Sparkdataframetransformationsparquetfile=spark.read.parquet(…)train_df=

parquetfile.withColumn(…)#Keras

APImodel=

Sequential().add(Convolution2D(32,3,3,activation='relu',input_shape=…))

\.add(MaxPooling2D(pool_size=(2,2)))

\.add(Flatten()).add(Dense(10,

activation='softmax')))#SparkML

pipelineEstimater=NNEstimater(model,CrossEntropyCriterion())

\.setLearningRate(0.003).setBatchSize(40).setMaxEpoch(5)

\.setFeaturesCol("image")nnModel=

estimater.fit(train_df)DistributedModel

ServingHDFS/S3KafkaFlumeKinesisTwitterSpoutAnalyticsZooModelSpoutBoltBoltBoltAnalyticsZooModelBoltBoltDistributedmodelservinginWebService,Flink,Kafka,Storm,

etc.PlainJavaorPythonAPI,withOpenVINOandDLBoost(VNNI)

supportAnalyticsZooUse

CasesComputerVisionBasedProductDefectDetectionin

Midea/en-us/articles/industrial-inspection-platform-in-midea-and-kuka-using-distributed-tensorflow-on-

analyticsNLPBasedCustomerServiceChatbotforMicrosoft

Azure/en-us/articles

溫馨提示

  • 1. 本站所有資源如無(wú)特殊說(shuō)明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請(qǐng)下載最新的WinRAR軟件解壓。
  • 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請(qǐng)聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶所有。
  • 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁(yè)內(nèi)容里面會(huì)有圖紙預(yù)覽,若沒有圖紙預(yù)覽就沒有圖紙。
  • 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
  • 5. 人人文庫(kù)網(wǎng)僅提供信息存儲(chǔ)空間,僅對(duì)用戶上傳內(nèi)容的表現(xiàn)方式做保護(hù)處理,對(duì)用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對(duì)任何下載內(nèi)容負(fù)責(zé)。
  • 6. 下載文件中如有侵權(quán)或不適當(dāng)內(nèi)容,請(qǐng)與我們聯(lián)系,我們立即糾正。
  • 7. 本站不保證下載資源的準(zhǔn)確性、安全性和完整性, 同時(shí)也不承擔(dān)用戶因使用這些下載資源對(duì)自己和他人造成任何形式的傷害或損失。

評(píng)論

0/150

提交評(píng)論