说明：  经验风险最小化原则(ERM)在样本有限时是不合理的,为了克服这一缺陷，1974年Vapnik&Chervonenkis提出了结构风险最小化原则(SRM)，即为了达到实际风险最小，需要同时使经验风险 和VC维h(置信范围)最小化。在深入研究经验风险 和实际风险 之间的关系即推广性的界后，得出了关于期望风险上界的估计，即对于给定的样本，当采用函数集 进行回归估计时，至少以概率 式成立
(Empirical risk minimization (ERM) is unreasonable when the sample is limited. In order to overcome this defect, Vapnik & chervonenkis proposed the structural risk minimization principle (SRM) in 1974, that is, in order to minimize the actual risk, both empirical risk and VC dimension H (confidence range) should be minimized. After a thorough study of the relationship between empirical risk and actual risk, i.e., the generalized bound, the estimation of the upper bound of expected risk is obtained, that is, for a given sample, when regression estimation is made by using function set, at least the probability formula holds)