ramsvm: Angle-based multicategory support vector machine with...

In bbeomjin/GBFSMSVM: Implementations of derivative-based variable selection on the multicategory support vector machines

Description

Fit an angle-based multicategory support vector machine with reinforced multicategory loss.

Usage

ramsvm(x = NULL, K = NULL, y, gamma = 0.5, lambda = 1,
	 kernel = c("linear", "gaussian"), kparam = 1, scale = FALSE, type = c("type1", "type2"), ...)

Arguments

x	A n x p data matrix, where n is the number of observations and p is the number of variables.
K	A kernel matrix computed from x and kernel function (optional). Given K, kernel and kparam are ignored.
y	A response vector with three and more labels.
gamma	The convex combination parameter of the loss function.
lambda	A regularization parameter to control a level of l_2-penalty.
kernel	A character string representing one of type of kernel.
kparam	A parameter needed for kernel.
scale	A logical indicating the variables to be scaled.
type	A type of optimization method for ramsvm. If type = "type1", intercepts are included in a regularization term. If type = "type2", intercepts are omitted in the regularization term and estimated by a linear programming
...	Other arguments that can be passed to ramsvm core function.

Value

An S3 object of class "ramsvm" containing the following slots

x	Input x.
K	Input K.
y	Input y.
y_name	The class labels of y.
gamma	The convex combination parameter of the loss function.
n_class	The number of classes.
lambda	Given regularization parameter.
kernel	Given type of kernel.
kparam	Given parameter for kernel.
cmat	The corresponding coefficients.
c0vec	The intercepts.
alpha	The Lagrange multipliers.
fit_class	Fitted class.
epsilon	Convergence tolerance in the ramsvm core algorithm.

Examples

require(dbvsmsvm)

bbeomjin/GBFSMSVM documentation built on Nov. 7, 2021, 10:20 p.m.