PSA在我所有的工作中，应该满意度能够排在第三或者第四的位置吧。但是它的发表却经历了不少曲折。最早投稿TGRS，结论是大改。一共三个审稿人，其中有一些相左的意见，我贴出来让大家看一下：

审稿人1

:This manuscript is not well written and its content does not seem to be very interesting.

审稿人2

:This is a very interesting paper that is clearly written in terms of material exposition.

审稿人3

：The manuscript is well written and the mathematical development clear

但是呢，审稿人1下面的意见显示出他可能没有能力去驾驭这篇文章的评审工作

(1) There is a conceptual mistake in the proposed method named PSA (principal Skewness Analysis). The manuscript declares that PSA uses skewness as nongaussian index. It seems to be a mistake. PSA should still actually use Kurtosis as nongaussian index in nature. Please check the analysis process carefully.

(2) The proposed method has same results as FastICA, and the difference is only in computational speed. According to the reviewer’s knowledge, the computational speed of FastICA algorithm has been very fast. It does not seem to be very significant that the proposed method (PSA) is only faster that FastICA.

其中第一条意见他竟然认为我PSA用的是峭度指标（Kurtosis），从这个意见已经可以明显看出，他根本没有能力来评审这篇文章。根据我的经验，遥感圈这样的审稿人还颇多。这真是一个令人苦恼的现象。第二条意见也比较扯，他说FastICA已经够快了，比它更快没啥意义的。我相信，由于高阶统计张量的引入而使得PSA相对于FastICA的优美便捷性，凡是能读懂文章的人，都会为之动容。可是对于这种没有能力被感动的审稿人，我只能无奈。

我当时心气比较高，心想，这么有原创性的思想竟然让我大改，我不投你了。于是换了一圈杂志，比如投过PAMI，结果被人家没送审一脚踢回来了，说我引用的文献基本都属于遥感圈，因此推荐我投TGRS；被PAMI拒审后，我心气持续增高，投了SCIENCE，结果不几天就来意见了，建议投TGRS；这会心气还没平顺，没你TGRS难道我文章还发表不了么，于是就投了RSE，结果RSE也拒审，且直接推荐投TGRS，说理论方法类的更适合IEEE社区。这让我稍微有点蒙了，于是也没怎么按照第一轮审稿人的意见改，就稍微糊弄一下又屈尊投TGRS了。过了不到两个月，审稿意见就来了，拒！！！而且主要的负面意见仍然来自第一轮第一个审稿人，我把大致意见贴出来：

(1) The proposed method has same results as FastICA, and the difference is only in computational speed. According to the reviewer’s knowledge, the computational speed of FastICA algorithm has been very fast. It does not seem to be very significant that the proposed method (PSA) is only faster than FastICA. Further, the proposed method (PSA) has same results as FastICA only when using skewness index, and FastICA has a wider scope than the proposed PSA because FastICA also can use the fourth and higher order moments.
(2) Further, unfortunately, authors did not investigate two inherent characteristics of hyperspectral data called Abundance Nonnegative Constraint (ANC) and Abundance Sum-to-one Constraint (ASC), both of which have clear physical meaning. It is very important to consider these two constraints in hyperspectral unmixing.

关于第一条意见，由于我在第一轮的回复中我驳斥了他的第一条意见，说我们PSA用的是偏度，而不是峭度。这个审稿人居然抓住了这一点，说FastICA可以用峭度指标或更高阶的统计指标，而我们PSA只能用偏度指标，所以PSA弱爆了。这意见真让人欲哭无泪，用协偏度（coskewness）张量叫PSA。同样的方法，我们用协峭度（cokurtosis）张量就是协峭度分析（PKA）了啊。更高阶的统计量，利用我们PSA的思想都不在话下啊。

关于第二条意见，这蛋扯的更没边没沿了，竟然拿混合像元分析中的和1约束和非负约束来扯。我和我的小伙伴们看到这条意见立刻集体呆若木鸡。我高光谱搞了10几年，这点破玩意还用你说？可是我们现在是在探讨特征提取方法，是在降维，不是在搞混合像元分解啊！！！这个审稿意见甚至让我万念俱灰，遥感圈中都是这种人的话，我以后发文章就困难了，因为稍微不控制一下，就超出他们知识范围了。

这个审稿人的意见最终左右了主编aplaza，最终给出了reject的结论。不过主编大人还算有点良心，推荐我们改投GRSL，结果四个审稿人全部绿灯通过，最终顺利发表。这四个哥们倒是令我意外了，他们都看懂文章，我实在不抱希望，这点从审稿意见也看的出来。好在发表了就行，就不深究了。