1.svm had a clear way of doing capacity control! such as L2 regularization,and different kernel function.
2.how to describ the capacity of svm?
3.an svm with a "narrow" kernel function can always learn the training set perfectly,but its generation error is controlled by the width of the kernel and the sparsity of the dual coefficients.
4.svm 的强对偶对应着数据可分的情况,如果数据不可分那么强对偶不再成立.