摘要：WeproposeaninexactproximalaugmentedLagrangianframeworkwithexplicitinnerproblemterminationruleforcompositeconvexoptimizationproblems.Weconsiderarbitrarylinearlyconvergentinnersolverincludinginparticularstochasticalgorithms,makingtheresultingframeworkmorescalablefacingtheever-increasingproblemdimension.Eachsubproblemissolvedinexactlywithanexplicitandself-adaptivestoppingcriterion,withoutrequiringtosetanaprioritargetaccuracy.Whentheprimalanddualdomainarebounded,ourmethodachievesthebestknowncomplexityboundsintermsofnumberofinnersolveriterations,respectivelyforthestronglyconvexandnon-stronglyconvexcase.Withouttheboundednessassumption,onlylogarithmtermsneedtobeadded.Withinthegeneralframeworkthatwepropose,wealsoobtainthefirstiterationcomplexityboundsunderrelativesmoothnessassumptiononthedifferentiablecomponentoftheobjectivefunction.Weshowthroughtheoreticalanalysisaswellasnumericalexperimentsthecomputationalspeedupachievedbytheuseofrandomizedinnersolversforlarge-scaleproblems.

Introduction：Theworkshopisthe8thmeetingoftheICCM(InternationalCongressofChineseMathematician)ConsortiumonCAM(ComputationalandAppliedMathematics).TheconsortiumhasthemissionofenhancingcollaborationamongChineseMathematiciansworkingonCAMandenhancingtheircollaborationwithinternationalexperts.The8thworkshopshallfocusonAI,BigDataandBlockChain,fromapplicationsofcomputationalandappliedmathematicsandasanenlargedscopeofappliedmathematics.ScientificCommittee：RaymondChan,TheChineseUniversityofHongKongXiaotieDeng,PekingUniversityXianfengGu,TheStateUniversityofNewYorkatStonyBrookYizhaoHou,CaliforniaInstituteofTechnologySenHu,USTCWen-WeiLin,NationalChiaoTungUniversityHongWan,NorthCarolinaStateUniversityZoupingXin,TheChineseUniversityofHongKongShing-TungYau,HarvardUniversityJunZhang,UniversityofMichiganJunZou,TheChineseUniversityofHongKongInvitedSpeakers:ChangwenChen,TheChineseUniversityofHongKong(Shenzhen)XiaotieDeng,PekingUniversityTiexiangLi,SoutheastUniversityAndreyLosev,MoscowHigherSchoolofEconomicsXiaokangMo,NorthValleyInnovationZhanPang,PurdueUniversityDonaldRubin,TsinghuaUniversityHongWan,NorthCarolinaStateUniversityDongmingWang,CNRS-FranceXinweiXie,ShanghaiHaijuBigdataIncChaopingXing,NanyangInstitueofTechnologyYanXu,USTCYongYang,ShanghaiHaijuBigdataIncJuyongZhang,USTCHongZhang,USTCJunZhang,UniversityofMichiganZeminZheng,USTCJunZou,TheChineseUniversityofHongKongLocalOrganizingCommittee：SenHu,USTCCanhongWen,USTCYanXu,USTCHongZhang,USTCJuyongZhang,USTCZeminZheng,USTCMinisymposiums:Minisymposium1:（AIandBigData）Minisymposium2:（BlockChain）Ifinterestedingivingatalkatoneofthemini-symposiumspleaseregisterandsubmitatitleandanabstract.ConferenceRegistrationhttp://abc.sen-cloud.com.cn（PleasecopytheURLandopenitinanewwebpage）Sponsors:The8thICCM-CAMAnnualWorkshop（全球华人数学家联盟-应用与计算数学学术会议）onAI,BigDataandBlockChainissupported/sponsoredby:USTC（中国科学技术大学）HefeiCityGovernment（合肥市政府）CSIAMActivityGrouponBlock-Chain(中国工业与应用数学学会区块链专业委员会）

第六届交叉科学与应用论坛--数据驱动研究的理论与方法探讨

计算机学院青年学术沙龙-安曼德副研究员学术报告会题目:Vehicleto-everything:StandardizationandChallenges时间:2019年12月12日16:00地点:西区电三楼632学术报告厅

香港科技大学易珂教授学术报告会题目:RandomSamplingonBigData:TechniquesandApplications时间:2019年12月12日10:00地点:西区电三楼632学术报告厅

报告人简介：李家康博士剑桥大学博士毕业，美国国家工程院院士，广东康云多维视觉智能科技有限公司联台创始人，CEO兼CTO。中国图像图形学学会第一位外籍会员，华南理工大学校外硕士生导师，广东省图像图形学学会计算机视觉专业委员会副主任，广东省图像图形学学会虚拟现实与智能交互专业委员会副主任。创办英国图形图像引擎公司，实现年收入1.4亿美金；个人拿下Motorola全球总订单5000万美金合同；发明专项70项。创立康云多维视觉智能科技有限公司，康云公司荣获2019年第八届中国创新创业大赛“人工智能与先进制造”成长组全国总冠军。报告摘要：Wealreadyknowthatsupervisedmodelscanperformextremelywellonnumeroustasksbuttheoutstandingaccuracyhasitsownprice.Nevertheless,AIresearchesdoagreatjobincreatingmoreandmorepowerfulmathematicalmodels,theoneswhoarefeedingthesemodelsoftenmisusethemhorriblyorthesemodelsevenaren’tmeanttomeettheexpectations.Algorithmsthatworkwellinresearchphasecanfailontotallyunexpectedproblems.Forexample,itturnedoutthatwidelyusedtodayconvolutionalneuralnetworks(CNNs)aresufferingfromadversarialattacksandtendtolearnnottheshapeofthevisualobjects,butrathertheirtexture(pictureandthelinkabove).Theproblem,probably,liesinthecoremathematicaloperationchosenforthemodel—convolutions,thatarenotrobustenoughastheyseemlike.GenerativeAIwilldemocratise3Dcontentcreationanddisplayinasimilarmannertowhathashappenedinothercreativedisciplinessuchasvideoproduction,photography,imageprocessingandgamedevelopment.AtLight&Magic,we’repassionateaboutbringingthesenew“automation”toolstolifeandlesseningthegapbetweenideaandcreation.WeusedGENERATIVEAI(embeddedinhardware&software)for3DSpatialRealityComputing.Providingultra-fastscanningandmappingsolutionsfor3Dmeasurement,documentation&realisation.

香港浸会大学徐建良教授学术报告会题目:TowardsSearchableBlockchain时间:2019年12月9日10:00地点:西区电三楼632学术报告厅

计算机学院青年学术沙龙--连德富研究员学术报告会题目:面向高效在线匹配的推荐系统时间:2019年12月5日16:00地点:西区电三楼632学术报告厅

报告人简介：陈文莉博士,新加坡南洋理工大学国立教育学院学习科学与科技学部副教授，博士生导师。长期致力于移动学习、计算机支持的协作学习、人工智能支持的学习分析等领域教学与研究。陈文莉博士主持了众多研究项目，尤其在科技增强科学教育领域。她是《JournalofComputersinEducation》（ESCI收录）的主编，并担任《InstructionalScience》（SSCI收录）、《AsiaPacificJournalofEducation》（SSCI收录）和《ResearchandPracticeinTechnologyEnhancedLearning》副主编，以及在《AsiaPacificEducationReview》（SSCI收录）担任顾问编辑。陈文莉博士是新加坡教育部教育技术类项目专家组成员，新加坡国家研究基金高等教育科技增强学习项目评审委员。她担任国际学习科学学会CSCL委员会主席，被亚太计算机教育应用协会授予“YoungResearcherLeader”奖项。

计算机学院青年学术沙龙-博士研究生李昭研究进展报告会题目:AutoSys：基于学习算法的云系统自动优化框架时间:2019年11月21日16:00地点:西区电三楼632学术报告厅