Improved Learning Rates of a Functional Lasso-type SVM with Sparse Multi-Kernel Representation Shaogao Lv