保序回归是寻找使训练集均方差最小的近似函数，它的优点是目标函数不要线性的。-The isotonic regression finds a non-decreasing approximation of a function while minimizing the mean squared error on the training data. The benefit of such a model is that it does not assume any form for the tar