ksmm: Kernel Support Matrix Machine

In kyuridata/ksmm: ksmm algorithm

Description

ksmm is used to train a kernel support matrix machine. It can be used to carry out classification.

Usage

	ksmm(x = NULL, y, K = NULL, x_dim, cost, kernel = c("linear", "rbf"), sigma = 1,
                maxit = 1e+4, epsilon = 5e-2, init_alpha = 1e-6, nCores = 1, ...)

Arguments

x	a input data matrix
y	a response vector with one label for each row / component of X. It must be 1 or -1
K	the calculated kernel matrix. If you don't input anything, it works automatically.
x_dim	the size of input matrix X
cost	cost of constraints violation
kernel	type of the kernel used in training. linear : linear kernel. rbf : Gaussian RBF kernel.
sigma	kernel width for the Radial Basis kernel function "rbf"
maxit	a maximum number of iteration for SMO algorithm
epsilon	a minimum value for optimization
init_alpha	initialization of alpha
nCores	the number of cores to use for parallel computing.
...	currently not used.

Details

This function is built by KSMM paper. Detailed theory is included in the KSMM paper.

Author(s)

Kyuri Park (based on Matlab code by Yunfei Ye)

References

Ye, Y. (2019). A nonlinear kernel support matrix machine. International Journal of Machine Learning and Cybernetics.

See Also

make_ksmm_kernel

Examples

require(ksmm)
data(nottingham)

X = as.matrix(nottingham[,-1])
y = ifelse(nottingham[,1] == 1, 1, -1)
train_x = X[1,]
train_y = y[1]

ksmm_fit = ksmm(train_x, train_y, c(200,200), 1, 1, 'rbf', 100, 1, 5e-2)
    	

kyuridata/ksmm documentation built on Dec. 21, 2021, 8:47 a.m.