Generalized Augmentation of Multiple Kernels

Kernel combination is meant to improve the performance of single kernels and avoid the difficulty of kernel selection. The most common way of combining kernels is to compute their weighted sum. Usually, the kernels are assumed to exist in independent empirical feature spaces and therefore were combined without considering their relationships. To take these relationships into consideration in kernel combination, we propose the generalized augmentation kernel which is extended by all the single kernels considering their correlations. The generalized augmentation kernel, unlike the weighted sum kernel, does not need to find out the weight of each kernel, and also would not suffer from information loss due to the average of kernels. In the experiments, we observe that the generalized augmentation kernel usually can achieve better performances than other combination methods that do not consider relationship between kernels.