这是一篇关于科研中如何选择数据测试集的文章，文章颠覆性的说明了我们在大多数实验中所谓的算法通用性，其实都在使用倾向我们结论的数据，而真正适合“所有”算法测试的数据集并没有得到验证。-This is an article about how to choose the scientific research data set of tests of the article, the article subversive to explain our in most of the so-called algorithm in the general, actually use tend to our conclusions data, and really suitable for "all" algorithm testing data JiBing not verified.