24小时热门版块排行榜

>论坛更新日志 (2189)
>虫友互识 (87)
>公派出国 (43)
>导师招生 (33)
>休闲灌水 (30)
>论文投稿 (28)
>基金申请 (24)
>考博 (24)
>硕博家园 (22)
>找工作 (18)
>教师之家 (16)
>文献求助 (16)
>考研 (13)
>论文道贺祈福 (11)
>博后之家 (8)
>专业外语 (7)

返回列表

seusu

木虫 (小有名气)

应助: 0 (幼儿园)
金币: 3143.9
帖子: 122
在线: 339.4小时
虫号: 340253

[交流] （外刊）审稿人反馈的意见让我感到很难办！所以求助

各位虫子好，好久没有发帖求助了。现在有机会了，呵呵（我倒是希望不要有这种机会，嘿嘿）。具体情况描述如下：

去年底投了一篇，论文中涉及采用FCM聚类算法确定T-S模型的前件（即模糊区划分）问题。

一审中，三个审稿人，第一个审稿人给出小修，第二、第三个审稿人给了大修。其中有两个意见让我觉得有点儿难以作答。如下：

Reviewer # 2

一审意见中的第一条：

1. It is unclear why clustering approach is used in building T-S models where linear functions are used as rule consequences. In my view, clustering may be more suitable for Mamdani-type fuzzy systems. For T-S models, the issue is to find some (fuzzy) regions with approximate linear behavior.

我是这么作答的：

It is true that clustering approach is usually applied to build Mnidani-type fuzzy systems. As the reviewer remarked, the issue for T-S models is to find some fuzzy regions with approximate linear behavior. Therefore, the clustering can also be used to find appreciated fuzzy partitioning for the data and then linear functional relationship between the inputs and output can be identified in each fuzzy partitioning using the membership degree matrix, see in [19 ~ 30], without claiming of completeness.

结果返回第二轮审稿意见，该审稿人给了小修，并只对以上问题不满意，意见如下：

The paper has been revised for improvements according to the review comments. But one point still seems unclear, why did you use FCM in building Sugeno-type fuzzy models? For instance, consider a set of points created roughly in terms of y=2x. Such data points constitute a fuzzy region for making a Sugeno-type fuzzy rule. But, with large variation of x, it would be difficult to detect this fuzzy region using FCM. It would be nice if authors could discuss such issues in the second-revised version.

虫友们，问题一如上，我该如何作答？求助！谢谢！！！

Reviewer #3

一审中的第一、第二条意见如下（即comment 1 和 comment 2）：

1. In the clustering based learning algorithms, the key issue is how to handle the input and output effectively in clustering. The traditional input only clustering or input-output clustering (the approach used in the paper) have been regarded less effective, as the input data and output data plays a completely different roles in identification of fuzzy systems. The more updated approach are the output based clustering approach, see the related papers in this topic in xxxxx (xxxxx是审稿人给出的一些推荐文献). So what is the best way to distinguish the output and input data in clustering is a point needed to further thinking to see where the traditional input-output clustering is the best approach here。

2. The used input-output clustering approach is a mismatch between the input-output clustering and T-S fuzzy system modeling. The data you use for clustering are input-output pair (xi, yi) where the system you model are (xi, Li(xi) ), where Li(.) is a linear function.

就以上两条意见，我作答如下：

1. First of all, we agree with the reviewer’s view point. Indeed, it is important to take into account more information of the output in clustering, see the above literatures suggested by the reviewer (These literatures are added in the revised paper so as to complete the reviews of the paper.). The input-output clustering is also important and investigated, see in [22-30], without claiming of completeness. In our current study, we consider the input-output clustering. In further, we will extend the xxxx(我论文提出的一种改版FCM算法) by considering more information of the output variable. Thanks the reviewer’s recommendation.

2.The FCM algorithm proposed in our paper is implemented on the input-output data pair (xi, yi), and it is used to establish T-S model for a given system when the data samples (xi, yi) can be observed from such system. In both two cases, data pair is (xi, yi). Li(.) is the linear behavior that should be identified in each fuzzy region. Therefore, there maybe exists no mismatch. Note that xi maybe include some information of output y, see the Examples 2 and 3 in the revised manuscript.

结果返回第二轮审稿意见，该审稿人也给了小修，并只对以上两个问题不满意，意见如下：

Although the revised manuscript has made some improvements in responding the reviewers’ comments, the main comments (comments 1 and 2) in my original review report have not been addressed properly. In particular, my original comment 2 is basically the same as Comment 1 by Reviewer 2. So at least the authors should make some real effort to address this issue!

个人感觉审稿人有点儿不耐烦啦。虽然我个人自我感觉对FCM算法及T-S模型比较熟悉，但还是觉得有点难以回答。请教虫子们。

针对以上问题，我不知如何回答妥贴。请教了！
编辑让我两周内返回修改稿我和修改意见。请虫子们给与帮助。不胜感谢。

回复此楼