针对不同重要性的样本，采用不同的惩罚因子进行逼近，在训练错误率和模型复杂度之间进行权衡。
Different penalty factors are assigned to the sample data of different importance, and the tradeoff can be determined between training errors and model complexity.
本文阐述了这两类风险的产生原因及其相互关系，提出了用户优先因子和风险权衡的理念。
This paper describes the cause of these two types of risks and the relationship between them. The concepts of customer-favored factor and risk balancing are also presented.
应用推荐