会议论文审稿意见

审稿人1：***Weakaspects:Commentstotheauthor:whataretheweakaspectsofthepaper?Thepresentationofthispapershouldbeimproved.Theexperimentshowsthat,secrecyrateisalmostthesameastraditionalmethodandwhatisthepromotionofusingNNtorelayselection.Beside,theefficiencyofthisschemeisnotclearlydemonstrated.***Recommendedchanges:Theauthorsshouldcompletetheexperimentbyaddingthecomputationcosts.Thereferencesareold,latestadvancementofthisissueshouldbementioned.TheauthorsmayputmorewordsonthemeaningofintroducingNNtorelayselection,because,theexperimentresultisnotthatelegant.弱势方面：应该改进本文的介绍。实验表明，保密率与传统方法几乎相同，推广使用NN进行中继选择的动机是什么。此外，该模型的有效性尚未明确证明。推荐的更改：1)作者应该通过添加计算成本来完成实验。2)参考文献陈旧，应提及此问题的最新进展。3)作者可能会更多地谈及将NN引入中继选择的意义，因为实验结果并不那么优雅。审稿人2：***Weakaspects:Commentstotheauthor:whataretheweakaspectsofthepaper?Ihavethefollowingconcerns.First,theEnglishwritingofthispapershouldbefurtherimproved.Forexample,what’sthemeaningof“perfectsecrecyperformance”?What’sthemeaningof“Comparedtotheconventionalrelayselectionscheme”?Theauthorsshoulddistinguish“comparedwith”from“comparedto”.Theorganizationofthispapershouldalsobeimproved,arelatedworksectionisneededtoreviewexistingworks,whichenablesreaderstoeasilyfollowthepaper.Second,theperformanceevaluationshouldbefurtherimproved,currentversionofanalysisseemstobesimpleandtrivial.InSectionIV.NumericalResults,theauthorsclaimthat“Itisobviousthatthefeedbackoverheadofproposedschemeishalfofthatoftheconventionalscheme”.Ithinkthisisthemaincontributionofthispaper,andtheauthorsshouldelaborateitinmoredetails(e.g.,providingthedetailedanalysisorexperimentresults).Theclaimof“obvious”islessrigorousandfailstoconvinceme.***Recommendedchanges:Recommendedchanges.Pleaseindicateanychangesthatshouldbemadetothepaperifaccepted.TheEnglishwritingofthispapershouldbeimproved.Theperformanceevaluationshouldbefurtherimproved.弱势方面：1)首先，本文的英文写作应进一步完善。例如，“完美安全速率”（perfectsecrecyperformance）的含义是什么？“与传统的继电器选择方案相比（Comparedtotheconventionalrelayselectionscheme）”是什么意思？还应该改进本文的组织，需要相关的工作部分来审查现有的作品，使读者能够轻松地按照论文进行。2)其次，绩效评估应（theperformanceevaluation）该进一步提高，目前的分析版本似乎简单而琐碎。在第四节。数值结果，作者声称“很明显，所提出的方案的反馈开销是传统方案的一半”。我认为这是本文的主要贡献，作者应该更详细地阐述它（例如，提供详细的分析或实验结果）。“显而易见”的主张不那么严谨，无法说服我。推荐的更改：1)应该改进本文的英文写作。2)绩效评估应进一步改进。审稿人3：***Weakaspects:Commentstotheauthor:whataretheweakaspectsofthepaper?However,somepartsofthispaperisconfusing.1.Theauthorsclaimthatnoliteratureshavereportedtheapplicationofmachinelearninginrelayselectionbytheknowledgeoftheauthors.Butsomesimilarideashavebeenproposedinrecentyears,suchasSankhe,Kunal,etal.MachinelearningbasedcooperativerelayselectioninvirtualMIMO.arXivpreprintarXiv:1506.01910(2015).2.Themachinelearningbasedrelayselectionisnotclear.Thetrainingdatasetandthetestingdatasetaresocalledlegitimatechannelcomplexmatrixandthewiretapchannelcomplexmatrix.Howthesedatacanbeobtained/generated?WhyitisimportantandsignificantastheinputoftheNN-basedapproachandwhatistheoutput?Thewholeprocessisnotclear.3.Theauthorsclaimthatthemachinelearningbasedapproachisbetterthantheconventionalapproach,butthenumericalresultsdonotshowtheconclusion.Theachievablesecrecyrateisalmostsameandtheonlydifferenceisthefeedbackoverheads.Onlyashortdiscussionisrelatedtofeedbackoverheadsandmoredetailsareneeded.4.Theauthorschoosetheachievablesecrecyrateastheperformancemetricbutthereisnoreferencethatwhyitcanbetheperformancemetric.***Recommendedchanges:Recommendedchanges.Pleaseindicateanychangesthatshouldbemadetothepaperifaccepted.Theauthorsshouldaddresstheweakaspects.弱势方面：1)近年来提出了一些类似的想法，例如Sankhe，Kunal，etal。“虚拟MIMO中基于机器学习的协作中继选择。”arXivpreprintarXiv：1506.01910（2015）。2)基于机器学习的继电器选择不明确。训练数据集和测试数据集被称为“合法信道复合矩阵和窃听信道复合矩阵”。如何获取/生成这些数据？为什么它作为基于NN的方法的输入是重要的和重要的，什么是输出？整个过程尚不清楚。3)作者声称基于机器学习的方法比传统方法更好，但数值结果没有显示结论。可实现的保密率几乎相同，唯一的区别是反馈开销。4)只有简短的讨论与反馈开销有关，需要更多细节。5)作者选择可达到的保密率作为绩效指标，但没有提及为什么它可以作为绩效指标。推荐的更改作者应该解决薄弱环节。