Multiple feature inputs support vector regression

While traditional support vector regression algorithm is based on a single kernel, in practice it is often desirable to use multiple kernels. Learning linear combinations of multiple kernels is an appealing strategy when the right choice of feature is unknown. A new support vector regression framework for multiple feature inputs learning problem is proposed here, which mapped the different feature inputs with different nonlinear functions into high dimensional feature space, and resolved as a semi-infinite linear program. Experimental results show that the proposed algorithm works well, and helps to improve the accuracy and the generalization for regression problems with multiple feature inputs.