
June 16, 2015 13:45 PSP Book - 9in x 6in 09-Yong-Ching-Lim-c09
References 307
multikernel algorithm to provide a more localised estimation, a
direct benefit from larger training sets, and also to assign tailored
kernel combinations to each region of the input space.
9.6 Discussion
After revisiting the standard support vector regression framework,
we have reviewed three approaches to higher-dimensional kernels
(HDK): complex-, quaternion-, and vector-valued kernels. For each
of these cases, we have discussed their theoretical requirements
and have illuminated how their feature spaces allow for the design
of kernel estimation algorithms. We have also shown ...