报告摘要：物理生物学是物理与生物相结合的一门交叉学科。本报告将举例介绍物理生物学目前的一些典型研究方法和内容，并简单介绍科大物理生物学相关实验室的情况。报告人简介：袁军华，现任中国科大物理系及合肥微尺度物质科学国家研究中心教授。1998年本科毕业于中科大；2005年获加州理工学院物理博士学位；2006-2012年在哈佛大学任博士后及Researchassociate。2012年通过国家青年千人计划回科大任教。2019年获国家自然科学基金委杰青项目资助。主要从事细菌运动的物理机制及微生物环境感知等方面的研究。

摘要：Technologyadvancementsindiagnosticimaging,smartsensing,andhealthinformationsystemshaveresultedinaBigDataenvironmentinhealthcare.Itisnowpossibletotrackeverypieceofinformationrelatedtoapatient\'scarecycleincludingdiagnosis,prognosis,treatment,caredelivery,andcontinuousmonitoring.ThisoffersagreatopportunityforPreciseMedicine.Ontheotherhand,thesizeandcomplexityofthedataoverwhelmthemodelingcapabilityofexistingstatisticalmethods.Inthistalk,Iwillpresenttwotopicsthattacklethedatasciencechallengesinmigrainediagnosis/subtypingandincarecoordination,respectively.Inthefirsttopic,aMultimodalityFactorMixtureModel(MFMM)isdevelopedtoenablemigrainesubtypediscoveryfrommultimodalityimagingdata.MFMMemploysanoveldoubly-penalizedformulationtoachievehierarchicalselectionofinformativeimagingmodalitiesandinformativefeatureswithineachmodality.Thisworkenablesrefinedclassificationofmigrainepatientsandsubtype-optimizedtreatment.Inthesecondtopic,IwillpresentthedevelopmentofaMulti-responseMulti-levelModel(M3)tofusemulti-sourcedatacollectedbytherecentlydevelopedNurseCareCoordinationInstrument(NCCI),inordertorevealhowcarecoordinationactivitiesareaffectedbynursetraining,workload,andtheirpracticeenvironment.Thisworkenablesdevelopmentofbestpracticestoimprovecarecoordinationandpatientoutcomes.Bothworkswillbedemonstratedbyreal-worlddataandcasestudies.主讲人简介：BingSiisanAssistantProfessorinSystemsScienceandIndustrialEngineeringatStateUniversityofNewYorkatBinghamton.ShereceivedherB.S.inMathematicsfromUniversityofScienceandTechnologyofChinaandanM.S.andaPh.D.inIndustrialEngineeringfromArizonaStateUniversityin2014and2018,respectively.Bing\'sresearchfocusesondevelopingdataanalyticsandstatisticallearningmethodologiestosupporthealthcaredecisionsindiagnosis,prognosis,treatment,andcaredelivery.HerresearchhasbeenappliedtoanumberofdiseasedomainsincludingAlzheimer\'sdisease,migraine,andtraumaticbraininjury.ShehasexperiencecollaboratingwithmedicalprofessionalsinMayoClinic,ASUSchoolofNursingandHealthInnovation,andthemedicalschoolatTechnicalUniversityofMunich,Germany.Sheisarecipientofmultipleawardsandscholarships,includingDean\'sDissertationAwardfromIraA.FultonSchoolsofEngineeringatASU,OutstandingEmergingFultonStudentOrganizationLeader(forservingasVPofINFORMSStudentChapter),GraceHopperScholarshipforWomeninComputing,IISEConferenceHealthcareStudentBestPaper3rdPlace,Quality&ProductivityResearchConferenceStudentScholarship,andASUCIDSEDoctoralFellowship.SheisamemberofIISE,INFORMS,andIEEE.

摘要：IRdualityisoneofthefascinatingphenomenasinquantumfieldtheory,inwhichtwoormoreQFTsenjoythesameIRphysics.IwillfocusonIRdualitiesofminimallySUSYfieldtheoriesin4,2,0dimensions,namely4dN=1Seibergduality,2dN=(0,2)triality,0dN=1quadrality.Therearecasesthatalarge(eveninfinite)numberofQFTsareconnectedbythesedualities,whichcanbeorganizedintoso-calleddualityweb.EngineeringtheseQFTsinstringtheorywithbranesprovidesnewperspectivesforstudyingthesedualitiesviageometry,whichmakesitpossibletoclassifyandcharacterizespacesofdualtheories.

报告人介绍：G．QWang现任中国电科院创新中心未来网络技术首席研究员，是业界资深网络技术专家，曾在加拿大北方电讯和华为美国研究院工作30余年，长期从事光纤网络、IP/MPLS网络、4G/5G无线网络和未来Information-Centric网络架构/协议技术及其应用的研发工作。G．QWang主要研究成果及所获荣誉包括67项美/欧授权专利、参与发表国际会议文章40余篇、IETF,IEEE等国际标准组织成员、华为“明日之星2015”获得者、加拿大北方电讯“杰出工程师NortelDMTS(DistinguishedMemberofTechnicalStaff)2008”获得者以及GMPLS网络技术发明人。SatellitenetworksareconsideredaninfrastructureofdatatransportforfutureB5G/6Gcommunication.ThistalkwillfocusonhowtouseVSTARroutercontext-labelforwardingtechnologytovirtualizesatellitecommunicationresourceandtoimplementserviceslices.ItwillalsointroducesomeVSTARdetailssuchasserviceworkflow,slicenamingandroutingtrustinessverification.

Abstract：Inthisintroductoryreport,wewillconcerntheMcShaneidentityanditsvariousgeneralizationsandapplications.WewillpayspecialattentiontoMirzakhani’scelebratedrecursiveformulafortheWeil-Peterssonvolumeofthemodulespaceofcurveswithmarkedpoints.

摘要：Let$kge1$beanintegerand$mu$bethehalfofa{itnonzero}integer.Thefollowingstatementsholdfortheellipticco-adjointorbitoftherealLiealgebra$mathfrak{so}(2,2k+2)$thatcorrespondstothedominantweight$(underbrace{|mu|,ldots,|mu|}_{k+1},mu)$.1.Thisorbitisdiffeomorphicto$mathrm{SO}_0(2,2k+2)/mathrm{U}(1,k+1)$.Asaresult,itispre-quantizable.2.Thisorbitisthetotalspaceofafiberbundlewithbasespacebeingthetotalcotangentspaceofthepuncturedeuclideanspaceofdimension$2k+1$andthefiberbeingdiffeomorphicto$mathrm{SO}(2n)/mathrm{U}(n)$.Asaresult,itadmitsacanonicalpolarization.3.ThegeometricquantizationofthisorbitwithitscanonicalpolarizationyieldstheHilbertofsquareintegrablesectionsofaHermitianvectorbundleoverthepuncturedEuclideanspaceindimension$2k+1$;moreover,thisHilbertspaceprovidesageometricrealizationfortheunitaryhighestweight$frak{so}(2,2k+2)$-modulewithhighestweight[(-k-|mu|,underbrace{|mu|,ldots,|mu|}_k,mu).]TheaboveresultsinLietheoryisobtainedfromthestudyofmagnetizedKeplermodelsindimension$2k+1$.

美国伊利诺伊大学厄巴纳-香槟分校李一川博士学术报告会题目:ADistributedSecondOrderMethodforConsensusOptimization时间:2020年1月6日15:30地点:西区电三楼632学术报告厅

摘要：Thispaperpresentsageneralframeworkforhigh-dimensionalnonlinearvariableselectionusingdeepneuralnetworksundertheframeworkofsupervisedlearning.Thenetworkarchitectureincludesbothaselectionlayerandapproximationlayers.Theproblemcanbecastasasparsity-constrainedoptimizationwithasparseparameterintheselectionlayerandotherparametersintheapproximationlayers.Thisproblemischallengingduetothesparseconstraintandthenonconvexoptimization.Weproposeanovelalgorithm,calledDeepFeatureSelection,toestimateboththesparseparameterandtheotherparameters.Theoretically,weestablishthealgorithmconvergenceandtheselectionconsistencywhentheobjectivefunctionhasaGeneralizedStableRestrictedHessian.Thisresultprovidestheoreticaljustificationsofourmethodandgeneralizesknownresultsforhigh-dimensionallinearvariableselection.Simulationsandrealdataanalysisareconductedtodemonstratethesuperiorperformanceofourmethod.

F5Networks李塑刚博士学术报告会题目:MobileSensingforLearningUserContext时间:2020年1月6日14:00地点:西区电三楼632学术报告厅

课程简介：K-stabilityisaconceptfromdifferentialgeometry,whichisoriginallydefinedbyTiantodeterminetheexistenceofKE-metriconagivenFanomanifold.Recently,algebraicmethodshavebeenintroducedintothisfieldtoadvancesitfastly.Inthisseriesoftalks,IwillindetailpresentthepictureofthesocalledalgebraicK-stability.