|
Return coefficients for SVs... Needs to be used directly with caution!
Summary on what is happening in libsvm internals with sv_coef
svm_model's sv_coef (especially) are "cleverly" packed into a matrix
nr_class - 1 x #SVs_total which stores
coefficients for
nr_class x (nr_class-1) / 2
binary classifiers' SV coefficients.
For classifier i-vs-j
General packing rule can be described as:
i-th row contains sv_coefficients for SVs of class i it took
in all i-vs-j or j-vs-i classifiers.
Another useful excerpt from svm.cpp is
// classifier (i,j): coefficients with
// i are in sv_coef[j-1][nz_start[i]...],
// j are in sv_coef[i][nz_start[j]...]
It can also be described as j-th column lists coefficients for SV # j which
belongs to some class C, which it took (if it was an SV, ie != 0)
in classifiers i vs C (iff i<C), or C vs i+1 (iff i>C)
This way no byte of storage is wasted but imho such setup is quite convolved
|