写在开始
早在八月初的时候,我投稿过一篇TEVC,那个时候也曾撰写过一篇详细的TEVC投稿指南:
在十月中旬的时候,我收到了这篇文章的一审意见,要求在45天内修改完成提交,至今天为止,整整45天的时间。这是我第一次回复审稿人的意见,所以在这45天的时间内,我丝毫不敢懈怠,卯足了劲对每一条审稿意见进行深入的分析与讨论。这篇文章一共有五位审稿人,其中有一人未返回意见(我仔细检查过邮件与投稿中心都没发现有这位审稿的意见),不同审稿人的意见不尽相同,有人缓和,有人严厉,有人询问,有人质疑,交织在一起,将我在平时学习过程中的不足具象化,同时也成为了我改进论文的基础。45天的时间里,我有过焦虑与不安,几乎周周都是9-10.5-7的模式,更甚有好几天七点半起床去改论文的情况,这导致我非常疲惫,一回到住的地方,就再也不想动,生日当天妈妈给我买的衣服,到现在为止都还没有打开...
这段时间,回想起来,虽然过得很累,但也感受到了什么叫光阴似箭,恍然就是一眨眼的功夫,上一秒还在和师弟师妹们侃:“努力拼搏奋斗45天,争取拿下TEVC”,下一秒就到了Revision的deadline。目前论文已经完成绝大部分的修改工作,由于导师比较忙,需要时间来审阅,所以我申请了三周的延期时间。今日忙里偷闲,正好想站在审稿人的角度,来分析分析他们会从哪些方面提出问题,我们该如何在论文第一版的时候该如何尽可能做到最好,希望对大家(特别是做理论方法的同学)有所帮助。
审稿意见的分析
审稿人意见总结
论文总共有五个审稿人,五个审稿人中四个审稿人给出了建设性的意见,我将这些意见进行了分类,涵盖了研究背景与动机、文献综述、实验、算法机制、以及语法和表达五个方面。所有审稿人的意见几乎都是一阵见血,而且对论文存在的问题把控极为准确,以至于多个审稿人会提出几乎完全一致的看法,这也揭示着“英雄所见略同”这亘古不变的规律,他们惺惺相惜的同时,就苦了我这个在学海中苦苦挣扎的年轻人...下面是分类的总结,诸如1-2此类的标识代表着审稿人1的第2条意见。
- 背景与动机 (6): 1-2, 1-4, 1-5, 2-1, 2-5, 4-7
- 综述 (4): 1-3, 2-4, 4-2, 5-1
- 实验 (8): 1-6, 1-8, 1-9, 2-6, 2-7, 4-5, 5-4, 5-5
- 机制 (7): 2-2, 2-3, 4-3, 4-4, 4-5, 5-2, 5-3
- 语法与表达 (4): 1-1, 1-7, 4-1, 4-8
可以很明显的看出来,专家们喜欢对你的动机提出质疑,同时也关注你的综述是不是写得好,也希望你补一补实验,然后再问问你的机制相关的问题,最后看看你的文章有没有什么语法和表达上问题。
背景与动机
部分审稿人质疑文章的动机,体现在对传统的方法存在的问题总结不够到位,(这篇论文中说的是传统的多种群方法盲目搜索,导致在同一个区域进行反复探索,而审稿人指出,现有的算法针对反复探索的问题已经有解决策略。针对此问题我转变了思路,将该问题定义为传统的多种群方法对问题结构的学习还不够,导致在已有的峰附近进行盲目搜索,同时还容易错过有希望的区域。)还有一些审稿人要求对背景进行补充。
- Reviewer 1-4: The statement that "the search process of a population involves a high level of blindness, leading to repeated search in the same region" is presented as a shortcoming of existing multipopulation methods and is part of the motivation for the current paper. However, this characterization is not accurate.
(传统多种群方法的缺点描述不准确,传统算法已经有策略解决在同一区域反复搜索的问题) - Reviewer 2-5: The authors mention that "the search range of each population does not adapt to the shape of the peaks, causing the loss of peaks that are tracked" as a shortcoming of existing multipopulation methods. I question this as a valid shortcoming because the shape of peaks in real-world dynamic optimization problems can be highly complex—often asymmetric, irregular, rotated, and stretched.
(质疑提出的问题的有效性,认为不对称、病态等复杂特征的峰无法有效学出其吸引域) - Reviewer 1-2: It is important to note that the class of problems considered in this paper pertains to tracking the moving optimum. However, dynamic optimization problems encompass a broader range, including scenarios where finding the global optimum is not the primary objective, such as robust optimization over time.
(动态优化应该从更广义的角度进行讨论) - Reviewer 2-1: Please discuss the optimization difficulties that exist in dynamic optimization problems solved by the proposed algorithm.
(讨论所解决的动态优化的难点) - Reviewer 4-7: Please describe how the proposed algorithm addresses the two limitations: 1) the search process of a population involves a high level of blindness, leading to repeated search in the same region. 2) The search range of each population does not adapt to the shape of the peaks, causing the loss of peaks that are tracked.
(讨论如何解决传统算法的缺点)
综述
综述这一部分很重要,其回顾了近年来在这个领域的研究进展。审稿人会从多个方面对综述进行审查,第一就是看有没有引用那些关键综述;第二就是看你是不是讨论了近年来的研究,引用了最新的文章;第三便是你综述的内容是否与你研究的小点息息相关,比如要做基于data-driven的动态优化,不能仅仅只综述传统的动态优化方法,必须有一部分内容要从data-driven的角度进行综述,紧密扣着你研究的小点。从下面的4条关于综述的意见里,能够大致覆盖上述说的三大点。
在期刊论文里的综述里,要短而精,首先要认识到那些领域内关键综述的作用,它们很详细地从各个方面阐述了该领域的进展,特别是那些最新的关键综述,引用它们能够拓宽本文的视野,而不是局限于自己的一亩三分地。要对本领域有实时前沿的了解,每周多涉猎有没有新的论文。
- Reviewer 1-3: The authors should reference these surveys to provide a broader context and acknowledge prior work in this area.
(缺少领域内关键综述的引用) - Reviewer 2-4: Indeed, a high quality paper should include some papers published in the last three years.
(缺少对近期的工作综述) - Reviewer 4-2: The authors should discuss some other kinds of multi-population-based methods.
(综述的视野太窄了) - Reviewer 5-1: The paper should include a more comprehensive review of recent methods in the field, especially the growing body of work on machine learning-based approaches in evolutionary and dynamic optimization.
(缺少对机器学习方法在动态优化中的应用的综述)
审稿意见:实验
实验这一部分更为重要,它直接关乎到提出的方法的有效性,这一部分也可能是提意见最多的地方,根据这篇文章的审稿意见,我大致将其分为三个方面,第一个方面是有没有与最新的算法进行比较,因为最新的算法一般都代表更强的性能(有部分可能是水的);第二个方面就是有没有与最新且具有挑战性的测试问题上进行测试,就有审稿人对我测试的问题有疑惑,认为我测试的那个问题太老了并且太简单了,需要测试最新的测试集来进一步验证;第三个方面是有没有落地,落地的概念就是测试实际问题,如果你的算法只能在benchmark上跑出很好的效果,但是一到实际问题中就不行了,那算法就还有很多可以完善的地方。
讲到这,我想插一句其他的讨论,我的师兄tqs问过我一个问题,别人都是问题来找解决办法,而我们是做算法来找问题,如果算法在找到的问题中的效果不如其他人,是不是能说明效果不行。我的看法是想做出一个通用的算法是几乎不可能的,我们不可能做到面面俱到,所有人都不能做到万无一失,师姐dyy也说过,得从问题出发,分析具体问题的特性,来有针对性的设计方法。所以说到一些实际问题中算法能力不行了,我们不要慌张,借助工具来分析这个实际问题到底为什么会让我们的算法能力下降了,搞清楚本质,再根据本质来设计其解决方法,不断完善算法,最后逐渐通用化,这才是大道。
对比最新的算法
- Reviewer 4-6: Some of the compared algorithms are too old and were published more than 10 years ago. Please use recently proposed algorithms in comparison. Also, the comparison algorithms should mainly be designed for DOPs.
(对比近期的算法)
测试最新的有挑战性的Benchmark
- Reviewer 1-9: The authors rely solely on the Moving Peaks Benchmark, which is overly simplified and may not reflect the complexities of real-world problems. It would be beneficial to test the algorithms on more complex and realistic benchmarks.
(测试更复杂的测试集) - Reviewer 2-7: The test problems are from Reference [39], a paper was published in 1999. What kind of characteristics of dynamic optimization problems are covered by this test problems? Additionally, please verify your algorithm on test problems designed in recent years.
(测试最新的测试集) - Reviewer 5-5: Scalability Tests: Conduct scalability tests on larger and higher-dimensional problems to demonstrate the robustness and adaptability of the proposed method across different problem scenarios.
(测试更复杂的测试集)
测试实际问题
- Reviewer 2-6: This paper involves too many parameters. Indeed, an optimization algorithm should be used to solve real-world problems. Additionally, please verify your algorithm on some real-world optimization problems.
(测试实际问题) - Reviewer 5-6: Scalability Tests: Test on Real-world Problems: Future work should apply SPMP to real-world dynamic optimization problems to validate its effectiveness and practical applicability beyond synthetic benchmarks.
(测试实际问题)
审稿意见:机制
机制这一块,就与你的算法息息相关,审稿人自己心里有一条线,他关注的内容如果没看懂,或者你没有提到,他就会发出提问,这时只需要好好的解释,然后根据解释对正文内容进行修改,使得你的方法更加清晰即可,有审稿人会认为你的算法参数需要自适应调整,如果手都调整不能很好的适用于各种问题,我建议一开始实现算法的时候,就把参数自适应放在脑子里,设计算法的时候就解决这个问题。我现在最担心的就是审稿人提出的参数需要自适应这一点意见,要做自适应,实验就完全得重做,我只是简单解释了一下算法有哪些参数,哪些参数自适应了,哪些参数结果敏感性分析了,并没有很好的去实现审稿人提出的基于学习的方法参数自适应的内容,所以还是有点慌。
- Reviewer 2-3: The proposed algorithm includes five main components, which is too complex. Indeed, a good optimization algorithm should be simple, efficient, and easily understood by readers.
(质疑算法机制太过于复杂) - Reviewer 4-3: I am confused about how to partition the solution space. Actually, I can understand how many subspaces can be divided, but how to divide the space (for example, D-dimensional) into N subspaces?
(询问解空间划分的方法) - Reviewer 4-5: In Algorithm 4, how to encounter the environmental changes in DOP is not described.
(未点明如何处理环境变化)
时间不够怎么办?
写信给主编,要求延期两周,注意一般不能延期超过两周,只有特殊情况才可以。写信的模板可以参考下面:
Dear Editor-in-Chief,
We are writing to you regarding our submitted article entitled "论文标题" (Manuscript ID: 论文ID), which is currently under revision.
We have revised the manuscript based on the reviewers' suggestions and comments and have completed approximately 80% of the revisions. As the deadline for submitting the revised manuscript is approaching, we are concerned that we may not be able to finalize the revisions in time. To ensure the quality and completeness of the revised version, we would like to kindly request a three-week extension of the deadline, until December 19th.
We sincerely appreciate your understanding and assistance in this matter and look forward to your advice at your earliest convenience. Thank you for your kind consideration.
Kind regards,
Mai Peng
写在最后
对于我如何回复的这些意见,在此已经不重要了,重要的是能不能从源头解决这些问题,从源头解决了问题,那审稿人就会对你的文章刮目相看,你就会少走很多弯路。最后,祝我二审顺利,祝大家成功吧!
3 条评论
会实现的mi
都会成的 Mike
祝大家都成功!!