简介:Thispaperproposesasemi-greedyframeworkforoptimizingmulti-joinqueriesinshared-nothingsystems.Theplangeneratedbytheframeworkcomprisesseveralpipelines,eachperformingseveraljoins.Theframeworkdeterminesthe“optimal”numberofjoinstobeperformedineachpipeline.Thedecisionsaremadebasedonthecostestimationoftheentireprocessingplan.Twoexistingoptimizationalgorithmsareextendedundertheframework.Ananalyticalmodelispresentedandusedtocomparethequalityofplansproducedbyeachoptimizationalgorithm.Ourstudyshowsthatthenewalgorithmsoutperformtheircounterpartsthatarenotextended.
简介:尽管物理问题可以发生在无界或实质地更大的域,流动问题的数字解决方案通常要求围住的域。在这种情况中,人工的边界是必要的。为由有限元素的海军司烧方程discretized的一个很好确定的人工的边界条件是“游手好闲之人”状况。为这的原因是这个条件在部分集成以后在变化明确的表达自动地出现的事实粘滞条款和压力坡度。这个条件是外面建立的大多数之一流动为海军司烧调节,但是进这个边界条件有很少的分析卓见。我们探讨存在的问题和有“一个方向性的游手好闲之人”的方程调节的海军司烧的弱解决方案的稳定性。与平常的“游手好闲之人”相对照调节状况提高了的这条边界稳定性性质。在情况中纯外面流动,条件等价于原来的,而在情况中在流动,消散的效果出现。我们显示出弱答案的存在并且由稳定、非稳定的流动的计算说明这个边界条件的效果。[从作者抽象]
简介:Appliedstatisticiansareoftenconfrontedwithstatisticalinferenceproblemsdealingwithsituationsinwhichthereappeartobenodata,ordataofonlylimitedusefulness.Forexample,whenattemptingtofindaconfidenceintervalforabinomialproportion,thesamplemaycontainnosuccesses.Suchascenariocouldbeencounteredwhenattemptingtoestimatetheincidenceofanextremelyraresideeffectassociatedwiththeadministrationofanewlydevelopeddrug.Inthisarticle,weuseexamplesforourexperiencesworkingwithscientificinvestigatorsanddescribeseveralscenariosinwhichthereappearedtobenousefuldata,ordataofonlylimitedusefulness.Wedescribethemethodswepreferforanalyzingthedatainthesesituationsandillustratetheirapplicationusingtheactualdatafromtheinvestigationsweparticipatedin.