| |
Multi-Kernel LP-SVM following Weston (PhD-thesis, chapter 6)
|
a=mksvm(kdict,hyperParam)
Generates a svm object with given hyperparameters.
Hyperparameters (with defaults)
kdict={} -- dictionary of kernels (a cell array of kernel objects)
C=Inf -- the soft margin C parameter
ridge=1e-13 -- a ridge on the kernel
balanced_ridge=0 -- for unbalanced data
Model
alpha -- the weights
b0 -- the threshold
Xsv -- the Support Vectors
Methods:
train, test, get_w
Example:
use dictionary of 3 kernels
[r a]=train(mksvm({kernel,kernel,kernel('rbf',2.5),kernel('rbf',2)}),toy)
loss(test(a,toy))
| |
Reference : Extensions to the Support Vector Approach, Chapter 6 |