Support vector machines (supported vectors MACHINE,SVM)

Svm:

1. Linear and nonlinear kernel functions;

2. Confidence interval structure in relation to neural network;

3. Training methods;

4.SVM LIGHT,LS-SVM;

5. VC Dimension

What is the difference between u-svc and c-svc?

In addition to the parameters, the two are basically the same.

C-svc c∈ (0,∞)

U-svc c∈[0,1]

C is a good feature, which is related to the ratio of support vectors to the ratio of training errors.

SVM solves the QR problem, the variable dimension = The number of training samples. So that the number of matrix elements is the number of training samples squared.

In SVM standard algorithm, it is necessary to solve the complicated QP problem, obtain the global optimal solution theoretically, and the majority of Lagrange multipliers are 0., so that the final decision only relies on a few support vectors, so that the solution shows sparse characteristics in the SVM method.

In LS-SVM, because the objective function of the optimization problem uses the error square term and the equality constraint condition, the QP problem of SVM is transformed into a set of linear equations, so that the Lagrange multiplier is proportional to the error term, the direct consequence is that the final decision function is related to all the samples. The sparse characteristics of the solution in the SVM method are also lost. However, the LS-SVM method can still achieve the sparsity of the solution to some extent by sorting the Lagrange multiplied by the final solution and using the "pruning" algorithm.

Support vector machines (supported vectors MACHINE,SVM)