简介:Thispaperstudiesonlineschedulingofjobswithkindreleasetimesonasinglemachine.Here“kindreleasetime”meansthatinonlinesetting,nojobscanbereleasedwhenthemachineisbusy.EachjobJhasakindreleasetimer(J)≥0,aprocessingtimep(J)>0andadeadlined(J)>0.Thegoalistodetermineaschedulewhichmaximizestotalprocessingtime(∑p(J)E(J))ortotalnumber(∑E(J))oftheacceptedjobs.Forthefirstobjectivefunction∑p(J)E(J),wefirstpresentalowerbound√2,andthenprovideanonlinealgorithmLEJwithacompetitiveratioof3.Thisisthefirstdeterministicalgorithmfortheproblemwithaconstantcompetitiveratio.Whenp(J)∈{1,k},k>1isarealnumber,wefirstpresentalowerboundminf(1+k)/k,2k/(1+k)g,andthenweshowthatLEJhasacompetitiveratioof1+┌k┐=k.Inparticular,whenalltheklengthjobshavetightdeadlines,wefirstpresentalowerboundmax{4=(2+k),1}(for∑p(J)E(J))and4/3(for∑E(J)).ThenweprovethatLEJis┌k┐/k-competitivefor∑p(J)E(J)andweprovideanonlinealgorithmHwithacompetitiveratioof2┌k┐/(┌k┐+1)forthesecondobjectivefunction∑E(J).
简介:Researchersoftensummarizetheirworkintheformofscientificposters.Postersprovideacoherentandefficientwaytoconveycoreideasexpressedinscientificpapers.Generatingagoodscientificposter,however,isacomplexandtime-consumingcognitivetask,sincesuchpostersneedtobereadable,informative,andvisuallyaesthetic.Inthispaper,forthefirsttime,westudythechallengingproblemoflearningtogeneratepostersfromscientificpapers.Tothisend,adata-drivenframework,whichutilizesgraphicalmodels,isproposed.Specifically,givencontenttodisplay,thekeyelementsofagoodposter,includingattributesofeachpanelandarrangementsofgraphicalelements,arelearnedandinferredfromdata.Duringtheinferencestage,themaximumaposterior(MAP)estimationframeworkisemployedtoincorporatesomedesignprinciples.Inordertobridgethegapbetweenpanelattributesandthecompositionwithineachpanel,wealsoproposearecursivepagesplittingalgorithmtogeneratethepanellayoutforaposter.Tolearnandvalidateourmodel,wecollectandreleaseanewbenchmarkdataset,calledNJU-FudanPaper-Posterdataset,whichconsistsofscientificpapersandcorrespondingposterswithexhaustivelylabelledpanelsandattributes.Qualitativeandquantitativeresultsindicatetheeffectivenessofourapproach.
简介:Utilizingdatafromcontrolledseismicsourcestoimagethesubsurfacestructuresandinvertthephysicalpropertiesofthesubsurfacemediaisamajoreffortinexplorationgeophysics.Denseseismicrecordswithhighsignal-to-noiseratio(SNR)andhighfidelityhelpsinproducinghighqualityimagingresults.Therefore,seismicdatadenoisingandmissingtracesreconstructionaresignificantforseismicdataprocessing.Traditionaldenoisingandinterpolationmethodsrarelyoccasionedrelyonnoiselevelestimations,thusrequiringheavymanualworktodealwithrecordsandtheselectionofoptimalparameters.Weproposeasimultaneousdenoisingandinterpolationmethodbasedondeeplearning.Fornoisyrecordswithmissingtraces,weadoptaniterativealternatingoptimizationstrategyandseparatetheobjectivefunctionofthedatarestoringproblemintotwosub-problems.Theseismicrecordscanbereconstructedbysolvingaleast-squareproblemandapplyingasetofpre-traineddenoisingmodelsalternativelyanditeratively.Wedemonstratethismethodwithsyntheticandfielddata.
简介:Alargenumberofdebrisflowdisasters(calledSeismicdebrisflows)wouldoccurafteranearthquake,whichcancauseagreatamountofdamage.UAVlow-altituderemotesensingtechnologyhasbecomeameansofquicklyobtainingdisasterinformationasithastheadvantageofconvenienceandtimeliness,butthespectralinformationoftheimageissoscarce,makingitdifficulttoaccuratelydetecttheinformationofearthquakedebrisflowdisasters.Basedontheaboveproblems,aseismicdebrisflowdetectionmethodbasedontransferlearning(TL)mechanismisproposed.Onthebasisoftheconstructedseismicdebrisflowdisasterdatabase,thefeaturesacquiredfromthetrainingoftheconvolutionalneuralnetwork(CNN)aretransferredtothedisasterinformationdetectionoftheseismicdebrisflow.Theautomaticdetectionofearthquakedebrisflowdisasterinformationisthencompleted,andtheresultsofobject-orientedseismicdebrisflowdisasterinformationdetectionarecomparedandanalyzedwiththedetectionresultssupportedbytransferlearning.
简介:AIM:Toinvestigateandcomparetheefficacyoftwomachine-learningtechnologieswithdeep-learning(DL)andsupportvectormachine(SVM)forthedetectionofbranchretinalveinocclusion(BRVO)usingultrawide-fieldfundusimages.METHODS:Thisstudyincluded237imagesfrom236patientswithBRVOwithamean±standarddeviationofage66.3±10.6yand229imagesfrom176non-BRVOhealthysubjectswithameanageof64.9±9.4y.Trainingwasconductedusingadeepconvolutionalneuralnetworkusingultrawide-fieldfundusimagestoconstructtheDLmodel.Thesensitivity,specificity,positivepredictivevalue(PPV),negativepredictivevalue(NPV)andareaunderthecurve(AUC)werecalculatedtocomparethediagnosticabilitiesoftheDLandSVMmodels.RESULTS:FortheDLmodel,thesensitivity,specificity,PPV,NPVandAUCfordiagnosingBRVOwas94.0%(95%CI:93.8%-98.8%),97.0%(95%CI:89.7%-96.4%),96.5%(95%CI:94.3%-98.7%),93.2%(95%CI:90.5%-96.0%)and0.976(95%CI:0.960-0.993),respectively.Incontrast,fortheSVMmodel,thesevalueswere80.5%(95%CI:77.8%-87.9%),84.3%(95%CI:75.8%-86.1%),83.5%(95%CI:78.4%-88.6%),75.2%(95%CI:72.1%-78.3%)and0.857(95%CI:0.811-0.903),respectively.TheDLmodeloutperformedtheSVMmodelinalltheaforementionedparameters(P<0.001).CONCLUSION:TheseresultsindicatethatthecombinationoftheDLmodelandultrawide-fieldfundusophthalmoscopymaydistinguishbetweenhealthyandBRVOeyeswithahighlevelofaccuracy.TheproposedcombinationmaybeusedforautomaticallydiagnosingBRVOinpatientsresidinginremoteareaslackingaccesstoanophthalmicmedicalcenter.
简介:现代生物学实验技术是生物科学专业研究生的必修课程之一。教学改革的目的是突出以教师引导为主线、以拓展学生的视野并与国际接轨为主体、通过充分运用课堂讲授和E-learning教学平台系统的各自优势,培养学生的创新意识和创新能力,培养学生综合运用基本的生物学实验技术并设计课题的能力。