Kernel Matrix-Based Heuristic Multiple Kernel Learning

Kernel theory is a demonstrated tool that has made its way into nearly all areas of machine learning. However, serious limitation kernel methods knowing which needed in practice. Multiple learning (MKL) an attempt to learn new tailored through the aggregation set valid known kernels. There are generally three approaches MKL: fixed rules, heuristics, and optimization. Optimization most popular; however, shortcoming optimization they tightly coupled with underlying objective function overfitting occurs. Herein, we take different approach MKL. Specifically, explore divergence measures on values matrices reproducing Hilbert space (RKHS). Experiments benchmark datasets computer vision feature task explosive hazard detection demonstrate effectiveness generalizability our proposed methods.