k-部排序算法的稳定性分析

Stability Analysis for k-partite Ranking Algorithms

研究在样本集中删除一个元素后k-部排序算法的稳定性。证明如果使用lγ作为排序亏损函数,那么算法在具有得分一致稳定的前提下也具有亏损一致稳定;并且如果对任意x,K(x,x)存在有限上界。那么通过最小化正则经验l-误差得到的排序算法具有良好的一致得分稳定。

Study the stability of k-partite ranking algorithms focus on deleting an element from the sample set.If A uniform score stability ranking algorithm use lγ as a loss function,then it has uniform loss stability;also,if for any x,K（x,x） has a limited upper bound,then the ranking algorithm by minimizing the regularization empirical l-error has good uniform score stability.