An Iterative Algorithm for Selecting the Parameters in Kernel Methods

Giving a certain training sample set, the learning efficiency almost depends on the kernel function in kernel methods. This inspires us to learn the kernel and the parameters. In the paper, a selecting parameter algorithm is proposed to improve the calculation efficiency. The normalized inner product matrix is the approximation target. And utilize the iterative method to calculate the optimal bandwidth. The defect detection efficiency can be greatly improved adopting the learned bandwidth. We applied the algorithm to detect the defects on tickets’ surface. The experimental results indicate that our sampling algorithm not only reduces the mistake rate but also shortens the detection time.