将p维的样本线性映射到k维空间上（k<p）。而且，这k个维度是相互正交（不相关）的。这k个方向就是主成分，并且是重新构造出来的，而不是简单地从原来p维中去除p-k维。(To map sample linear P dimension to K dimensional space (k<p). Moreover, the K dimensions are orthogonal to each other (not related). The K direction is the main component, and is re constructed, rather than simply from the original P dimension P-K dimension removal.)