《人工智能-数据科学比赛汇总 20196.docx》由会员分享，可在线阅读，更多相关《人工智能-数据科学比赛汇总 20196.docx（32页珍藏版）》请在taowenge.com淘文阁网|工程机械CAD图纸|机械工程制图|CAD装配图下载|SolidWorks_CaTia_CAD_UG_PROE_设计图分享下载上搜索。
1、人工智能/数据科学比赛汇总2019.6Note:ARIEL,amissiontomakethefirstlarge-scalesurveyofexoplanetatmospheres,haslaunchedaglobalcompetitionseriestofindinnovativesolutionsfortheinterpretationandanalysisofexoplanetdata.Youcanfindourpressrelease.ThefirstARIELDataChallengeinvitesprofessionalandamateurdatascientistsaround
2、theworldtouseMachineLearning(ML)toremovenoisefromexoplanetobservationscausedbystarspotsandbyinstrumentation.AsecondARIELDataChallengethatfocusesontheretrievalofspectrafromsimulationsofcloudyandcloud-freesuper-Earthandhot-Jupiterdatawasalsolaunchedtoday.Afurtherdataanalysischallengetocreatepipelinesf
3、orfaster,moreeffectiveprocessingoftherawdatagatheredbythemissionwillbelaunchedinJune.EntryDeadline:iFLYTEKAI开发者大赛5月21日-10月14日,2019/Hostby/Prize:100万RMBNote:iFLYTEKAI开发者大赛是由科大讯飞提议的顶尖人工智能竞赛平台会聚产学研各界力量面向全球开发者提议数据算法及创新应用类挑战推动人工智能前沿科学研究以及创新成果转化培育人工智能产业人才助力人工智能生态建立。2019年度第二届iFLYTEKAI开发者大赛将继续开放科大讯飞优质大数据资源及
4、人工智能核心技术面向全球开发者提议数据算法及创新应用类挑战。阿尔茨海默综合症预测挑战赛:基于老年度人在特定图片描绘任务中产生的语音给定语音数据中提取出的声学特征、主被试对话的切分信息、人工文本转写结果和对应的认知标签建立2分类模型预测认知标签正常或者认知障碍。挪动广告反欺诈算法挑战赛:挪动广告反欺诈需要强大的数据作为支撑本次大赛提供了讯飞AI营销云海量的现网流量数据作为训练样本参赛选手需基于提供的样本构建模型预测流量作弊与否。大数据应用分类标注挑战赛:选手基于提供的应用二级分类标签和假设干随机应用标注样本实现应用分类标注算法每个应用一个标签以应用最主要属性对应的标签为该应用的标签。工程机械核心
5、部件寿命预测挑战赛:由中科云谷科技有限公司提供某类工程机械设备的核心耗损性部件的工作数据包括部件工作时长、转速、温度、电压、电流等多类工况数据。祈望参赛者利用大数据分析、机器学习、深度学习等方法提取适宜的特征、建立适宜的寿命预测模型预测核心耗损性部件的剩余寿命。EntryDeadline:PredictingMolecularPropertiesNow-August28,2019/Hostby/Prize:$30,000Note:Canyoumeasurethemagneticinteractionsbetweenapairofatoms?EntryDeadline:ChallengeonDe
6、epLearningbasedLoopFilterforVideoCodingMay,25th-May,31st,2018/Hostby/Prize:NaNNote:Theparticipantsareencouragedtoinvestigateneuralnetworkbasedmethods(especiallyconvolutionalneuralnetworks)withdifferentnetworkstructures,inahopeofachievingthebestqualitywithlightestnetworkconfigurationforagoodtradeoffo
7、fefficiencyandcomplexity.EntryDeadline:The2ndLarge-scaleVideoObjectSegmentationChallengeMay.20-Sep.52019/Hostby/Prize:NaNNote:Asacontinuousefforttopushforwardtheresearchonvideoobjectsegmentationtasks,weplantohostasecondworkshopwithachallengebasedontheYouTube-VOSdataset,targetingatmorediversifiedprob
8、lemsettings,i.e.,weplantoprovidetwochallengetracksinthisworkshop.Track1:VideoObjectSegmentationTrack2:VideoInstanceSegmentationEntryDeadline:OpenEDSChallengeMay3-Sep16,2019/Hostby/Prize:$13,000USDx2Note:Intheabsenceofaccurategazelabels,weproposetoadvancethestateoftheartbycarefullydesigningtwochallen
9、gesthatcombinehumanannotationofeyefeatureswithunlabeleddata.Thesechallengesfocusondeeperunderstandingofthedistributionunderlyinghumaneyestate.WeinviteMLandCVresearchersforparticipation.Track-1SemanticSegmentationchallengeTrack-2SyntheticEyeGenerationchallengeEntryDeadline:Exoplanetimagingdatachallen
10、geMay16th-Sep16th,2019/Hostby/Prize:NaNNote:Thiscompetitioniscomposedoftwosub-challengesfocusingonthetwomostwidelyusedobservingtechniques:pupiltracking(angulardifferentialimaging,ADI)andmulti-spectralimagingcombinedwithpupiltracking(multi-channelspectraldifferentialimaging,ADImSDI).EntryDeadline:Pek
11、ingUniversityInternationalCompetitiononOcularDiseaseIntelligentRecognition(ODIR-2019)May18-Sep25,2019/Hostby/Prize:10,00,000RMB(140,000USD)Note:北京大学智慧之眼国际眼科疾病智能识别竞赛TheSGwillprovideparticipantswith5,000structureddesensitizedophthalmologicimagesetofpatientsage,sex,binocularcolorfundusphotosanddoctorsd
12、iagnosticreport.上工医信将为参赛者提供5000组包含患者的性别、年度龄、双眼彩色眼底照片以及医生印象报告等的构造化脱敏后眼科的数据集。Thepurposeofthischallengeistocompareapproachesofophthalmicdiseaseclassificationincolorfundusimages.Participantwillhavetosubmitclassificationresultsofeightcategoriesforallthetestingdata.Foreverycategory,aclassificationprobabil
13、ity(valuefrom0.0to1.0)denotesriskofapatientdiagnosedwithcorrespondingcategory.该竞赛的目的是比拟基于彩色眼底图像进展眼科疾病分类的不同方法。介入者必须提交所有测试数据集的八个类别的分类结果。对于每个类别分类概率值从0.0到1.0表示患者被诊断为具有相应类别的可能性/风险。EntryDeadline:The2ndChina(Hengqin)InternationalUniversityQuantitativeFinanceCompetition2019-04-19至2020-03-21/Hostby/Prize:140
14、万Note:第二届中国横琴国际高校量化金融大赛参赛要求参赛者应根据题目要求完成一篇包括量化金融策略原理、模型的假设、建立以及求解、计算方法的设计、分析以及检验、模型的改良等方面的书面报告即答卷并在规定竞赛期间内将参赛策略的市场运行进展模拟仿真竞赛。根据参赛策略的测试结果包括样本内以及样本外的收益程度及市场风险防范的效果等统一指标打分评比以市场的标准来决定优劣评价策略的回测以及实盘模拟表现同时考虑策略逻辑的稳健性以及创新性。竞赛评奖以策略的合理性、建模的创新性、测试策略的市场适应性及收益风险程度等结果为主要标准。RequirementsParticipantsshouldwriteareport
15、coveringquantitativefinancialstrategytheories1)Modeltheoreticalhypothesisanddescriptionofquantitivemodel2)Dataanalysis3)Strategybacktestingresultsandperformanceanalysis.Accordingtotherequirementsofthecompetition,participantsstrategieswillbebacktestedandpapertradedduringtherequiredperiod.Evaluationan
16、dscoringwillbaseonunifiedmeasurementsincludingreturn,volatility,maxdrawdownofthestrategiesandsoon.Thedeterminationofmeritsandevaluationofstrategybacktestandpapertradingperformancewillbemadeaccordingtomarketstandards,whiletherobustnessandinnovationofthestrategiclogicwillalsobetakenintoconsideration.K
17、eycriteriawillincludetherationalityofthestrategy,thecreativenessofthemodel,themarketadaptabilityofthetestingstrategyandthelevelofreturnandrisk.EntryDeadline:VisualDomainAdaptationChallenge(VisDA-2019)April9-Sept.27,2019/Hostby/Prize:NaNNote:Wearepleasedtoannouncethe2019VisualDomainAdaptation(VisDA20
18、19)Challenge!Itiswellknownthatthesuccessofmachinelearningmethodsonvisualrecognitiontasksishighlydependentonaccesstolargelabeleddatasets.Unfortunately,performanceoftendropssignificantlywhenthemodelispresentedwithdatafromanewdeploymentdomainwhichitdidnotseeintraining,aproblemknownasdatasetshift.TheVis
19、DAchallengeaimstotestdomainadaptationmethodsabilitytotransfersourceknowledgeandadaptittonoveltargetdomains.Thischallengeincludestwotracks:Multi-SourceDomainAdaptationChallengeSemi-SupervisedDomainAdaptationEntryDeadline:AlchemyContest5/22-9/30,2019/Hostby/Prize:total100,000RMBNote:TheTencentQuantumL
20、abhasrecentlyintroducedanewmoleculardataset,calledAlchemy,tofacilitatethedevelopmentofnewmachinelearningmodelsusefulforchemistryandmaterialsscience.Thedatasetlists12quantummechanicalpropertiesof130,000organicmoleculescomprisingupto12heavyatoms(C,N,O,S,FandCl),sampledfromthedatabase.Thesepropertiesha
21、vebeencalculatedusingtheopen-sourcecomputationalchemistryprogramPython-basedSimulationofChemistryFramework().EntryDeadline:MicroNetChallengeNeurIPS2019June1,2018-Dec13,2019/Hostby/Prize:NaNNote:Thecompetitionconsistsofthreedifferenttasks.Contestantsarefreetosubmitentriesforone,two,orallthreetasks.Co
22、ntestantsareallowedtoenteruptothreemodelsforeachtask,butwillberankedaccordingtotheirtopentryineachtask.Entriescanonlybetrainedonthetrainingdataforthetasktheyareenteredin.Nopre-training,oruseofauxiliarydataisallowed.ImageNetClassificationThedefactostandarddatasetforimageclassification.Thedatasetiscom
23、posedof1,281,167trainingimagesand50,000developmentimages.Entriesarerequiredtoachieve75%top-1accuracyonthepublictestset.CIFAR-100ClassificationAwidelypopularimageclassificationdatasetofsmallimages.Thedatasetiscomposedof50,000trainingimagesand10,000developmentimages.Entriesarerequiredtoachieve80%top-1
24、accuracyonthetestset.WikiText-103LanguageModelingAlanguagemodelingdatasetthatemphasizeslong-termdependencies.Entrieswillperformthestandardlanguagemodelingtask,predictingthenexttokenfromthecurrentone.Thedatasetiscomposedof103milliontrainingwords,217thousanddevelopmentwords,and245thousandtestingwords.
25、Entriesshouldusethestandardword-levelvocabularyof267,735tokens.Entriesarerequiredtoachieveaword-levelperplexitybelow35onthetestset.EntryDeadline:EndoscopicVisionChallenge2019June5-Oct13,2019/Hostby/Prize:NaNNote:AsavisionCAIchallengeatMICCAI,ouraimistoprovideaformalframeworkforevaluatingthecurrentst
26、ateoftheart,gatherresearchersinthefieldandprovidehighqualitydatawithprotocolsforvalidatingendoscopicvisionalgorithms.EndoVis2019Sub-challenges:SurgicalWorkflowandSkillAnalysisStereoCorrespondenceandReconstructionofEndoscopicDataEntryDeadline:GraphGolf:TheOrder/degreeProblemCompetition05-1311-26,2019
27、/Hostby/Prize:NaNNote:Findagraphthathassmallestdiameteraverageshortestpathlengthgivenanorderandadegree.GraphGolfisaninternationalcompetitionoftheorder/degreeproblemsince2021.Itisconductedwiththegoalofmakingacatalogofsmallest-diametergraphsforeveryorder/degreepair.Anyoneintheworldcantakepartinthecomp
28、etitionbysubmittingagraph.OutstandingauthorsareawardedinCANDAR2019,aninternationalconferenceheldinNagasaki,Japan,inNovember2019.EntryDeadline:TheAnimal-AIOlympicsApril-December2019/Hostby/Prize:$10,000Note:基于UnityMLAgentsToolkit的动物认知-AI挑战ThiscompetitionpitsourbestAIapproachesagainsttheanimalkingdomt
29、odetermineifthegreatsuccessesofAIarenowreadytocompetewiththegreatsuccessesofevolutionattheirowngame.EntryDeadline:GeopoliticalForecastingGFChallenge2April4,2018-Feb.1,2020/Hostby/Prize:$250,000Note:Solvers,whetherindividualsorteams,willcreateinnovativesolutionsandmethodstoproduceforecaststoasetofmor
30、ethan300questionsreferredtoasIndividualForecastingProblems(IFPs),releasedregularlyoverthecourseofthenine-monthChallenge.EntryDeadline:ModaNetFashionUnderstandingChallengeOct1,2018-Dec11,2019/Hostby/Prize:NaNNote:Inthischallenge,weevaluatemodelperformanceforthreetasks,objectdetection,semanticsegmenta
31、tionandinstancesegmentation.Youcanparticipatealltasksoranyoneofthembychoosingwhichresultstobeincludedinyoursubmission.EntryDeadline:二分类算法提供银行精准营销解决方案|练习赛2018年度12月29日-2019年度12月29日/Hostby/Prize:NaNNote:本练习赛的数据选自UCI机器学习库中的银行营销数据集(BankMarketingDataSet)EntryDeadline:SPIE-AAPM-NCIBreastPathQ:CancerCellula
32、rityChallenge2019Oct.15,2018-Dec.31,2019/Hostby/Prize:NaNNote:Participantswillbetaskedtodevelopanautomatedmethodforanalyzinghistologypatchesextractedfromwholeslideimagesandassignascorereflectingcancercellularityineach.EntryDeadline:Optimizingwell-beingatworkJan.1,2019-Jan.1,2020/Hostby/Prize:NaNNote
33、:Thischallengeproposestodevelopmachinelearningbasedapproachessoastopredictindividualscomfortmodelusingseveraltimeseriesofenvironmentaldataobtainedfromsensorsinalargebuilding.Theobjectiveistolearnaclassifierthatusesthesetimeseriesasinputstopredicttheassociatedcomfortclasscomputedasanaverageofthecomfo
34、rtclassesofallindividualsinthebuilding,assumedtoexperiencethesameenvironmentalconditions.EntryDeadline:Drug-relatedquestionsclassificationJan.1,2019-Jan.1,2020/Hostby/Prize:NaNNote:ThegoalofPososchallengeistopredictforeachquestiontheassociatedintent.EntryDeadline:DetectingbreastcancermetastasesJan.1
35、,2019-Jan.1,2020/Hostby/Prize:NaNNote:ThechallengeproposedbyOwkinisaweakly-supervisedbinaryclassificationproblem:predictwhetherapatienthasanymetastaseinitslymphnodeornot,givenitsslide.EntryDeadline:BuildingClaimPredictionJan.1,2019-Jan.1,2020/Hostby/Prize:NaNNote:Thegoalofthechallengeistopredictifab
36、uildingwillhaveaninsuranceclaimduringacertainperiod.Youwillhavetopredictaprobabilityofhavingatleastoneclaimovertheinsuredperiodofabuilding.EntryDeadline:CracktheneuralcodeofthebrainJan.1,2019-Jan.1,2020/Hostby/Prize:NaNNote:Thechallengegoalistoclassifythebrainactivitystateofananimalbasedonspikingact
37、ivitypatternsofitsindividualneurons.EntryDeadline:PredictionofSharperatioforblendsofquantitativestrategiesJan.1,2019-Jan.1,2020/Hostby/Prize:NaNNote:TheproblemisapredictionchallengethataimsathelpingtheCompanytobuildanoptimalblendofquantitativestrategies,givenasetofsuchstrategies.EntryDeadline:Histor
38、icalconsumptionregressionforelectricitysupplypricingJan.1,2019-Jan.1,2020/Hostby/Prize:NaNNote:Thegoalofthechallengeistopredict,basedontheanalysisofthecorrelationofayearofconsumptionandweathertrainingdata,theelectricityconsumptionoftwogivensitesforatestyear.EntryDeadline:Predictbraindeepsleepslowosc
39、illationJan.1,2019-Jan.1,2020/Hostby/Prize:NaNNote:Inthisdataset,wetrytopredictwhetherornotaslowoscillationwillbefollowedbyanotheroneinshamcondition,i.e.withoutanystimulation.EntryDeadline:SpatiotemporalPM10concentrationpredictionJan.1,2019-Jan.1,2020/Hostby/Prize:NaNNote:Inordertoprovideairqualityf
40、orecasts,PlumeLabshasbuiltauniquedatabasewithreadingscollectedbymonitoringstationsallovertheworld.TheproblemwesubmitconsistsinpredictingthePM10readingsofsomeairqualitymonitoringstationsusingthereadingsprovidedbythemonitoringstationsnearbyaswellasurbanfeatures.EntryDeadline:DynamicProfileForecastingJ
41、an.1,2019-Jan.1,2020/Hostby/Prize:NaNNote:Thischallengeisaboutforecastingdynamicprofilesvaluesfromtheirpastvaluesandallthecomponentsof.EntryDeadline:Solve2x2x2RubikscubeJan.1,2019-Jan.1,2020/Hostby/Prize:NaNNote:ThegoalistodesignanautomaticRubiksanalyzerthatestimatesthecurrentlengthoftheshortestpath
42、tothesolution.EntryDeadline:Exoticpricingwithmultidimensionalnon-linearinterpolationJan.1,2019-Jan.1,2020/Hostby/Prize:NaNNote:Thepurposeofthechallengeistouseatrainingsetof1millionpricestolearnhowtopriceaspecifictypeofinstrumentsdescribedby23parametersbynonlinearinterpolationontheseprices.EntryDeadl
43、ine:ScreeningandDiagnosisofesophagealcancerfromin-vivomicroscopyimagesJan.1,2019-Jan.1,2020/Hostby/Prize:NaNNote:Thegoalofthischallengeistobuildanimageclassifiertoassistphysiciansinthescreeninganddiagnosisofesophagealcancer.EntryDeadline:PredictionofdailystockmovementsontheUSmarketJan.1,2019-Jan.1,2
44、020/Hostby/Prize:NaNNote:Thegoalofthischallengeistopredictthesignofthereturns(pricechangeoversometimeinterval)attheendofabout700daysforabout700stocks.EntryDeadline:MEMENTO:MRIWhiteMatterReconstructionMarch7,2019-March4,2020/Hostby/Prize:NaNNote:Thiswillbea2-yearchallenge.Weaimtohost3sub-challengesev
45、aluatingourcurrentabilityto:(1)predictunseensignal(signalrepresentation;)(2)estimatemicrostructuralmeasures(signalmodeling;)(3)evaluatesensitivityandspecificityofpotentialbiomarkers(biomarkerevaluation;).EntryDeadline:PropensitytoFundMortgages25APR2019-6JUN2019/Hostby/Prize:$10000Note:Developamodelt
46、opredict,givenmortgageapplicationinformation,whetherthemortgagewillbefundedornot.Topredictwhetheramortgagewillbefundedusingonlythisapplicationdata,certainleadingfactorsdrivingtheloansultimatestatuswillbeidentified.Solverswilldiscoverthespecificaspectsofthedatasetthathavethegreatestimpact,andbuildamo
47、delbasedonthisinformation.EntryDeadline:IdentifyCharactersfromProductImages12MAY2019-9JUL2019/Hostby/Prize:NaNNote:Identifythecharactersfromproductimagefromalistof42possiblevalues.Whileusingmachinelearningtoperformimagerecognitioniscurrentlyoneofthemostpopularusecases,insomecases,theexistinglarge-sc
48、alemodelsaretoobroadtobeeffectiveforspecificbusinessusecases.Inthiscontestwewilluseadatadrivenapproachtoidentifythe“charactersinanimage(productimages).EntryDeadline:KiTS19ChallengeMarch15-August2,2019/Hostby/Prize:NaNNote:Thegoalofthischallengeistoacceleratethedevelopmentofreliablekidneyandkidneytumorsemanticsegmentationmethodologies.EntryDeadline:PAIP2019ChallengeApril15-September2,2019/Hostby/Prize:NaNNote:Thegoalofthechallengeistoevaluatenewandexistingalgorithmsf