[−][src]Function linfa_svm::SVClassify::fit_nu
pub fn fit_nu<'a, A: Float>(
params: SolverParams<A>,
kernel: &'a LinfaKernel<ArrayView2<'a, A>>,
targets: &'a [bool],
nu: A
) -> Svm<'a, A, Pr>
Support Vector Classification with Nu-penalizing term
This methods solves a binary SVC problem with a penalizing parameter nu between (0, 1). The dual problem has the form
ⓘ
min_a 1/2*a^tQ a s.t. y^t a = 0, 0 <= a_i <= 1/l, e^t a > nu
with Q_ij = y_i y_j K(x_i, x_j)
the kernel matrix.
Parameters
params
- Solver parameters (threshold etc.)kernel
- the kernel matrixQ
targets
- the ground truth targetsy_i
nu
- Nu penalizing term