Multiple Kernel Learning

Multiple Kernel Learning (MKL) has been proposed as an approach to simultaneously learn the kernel weights and the associated decision function in supervised learning settings.

In MKL the kernel

K

can be considered as a linear combination of

M

basis kernels

K(x, x') = \sum_1^M \beta_mK_m (x,x') \\ \beta_m \ge 0 , \sum_1^M \beta_m = 1

The solution of the learning problem for kernel methods can be written as

f(x) = \sum_1^N \alpha_i K(x_i, x) + b

In MKL we learn both coefficients

\beta_m

and

\alpha_i

in a single optimization problem.

www.cse.msu.edu

https://www.cse.msu.edu/~cse902/S14/ppt/MKL_Feb2014.pdf

Multiple Kernel Learning

Recommendations