svm.admm
is a simple function for solving largescale regularized linear/nonlinear
classification by using ADMM and IADMM algorithms. This function provides
linear L2regularized primal classification (both ADMM and IADMM are available),
kernel L2regularized dual classification (IADMM) as well as L1regularized primal
classification (both ADMM and IADMM are available). The training of the models perform well
practice.
1 2 3 
x.tr 
a n*p data matrix. Each row stands for an example (sample, point) and each column stands for a dimension (feature, variable). 
y.tr 
a nlength vector. The values correspond to class labels. 
type 

kernel 
the kernel used in training and predicting when

sigma 
The inverse kernel width used by the Gaussian. 
degree 
The degree of the polynomial kernel function. This has to be an positive integer. 
scale 
The scaling parameter of the polynomial kernel is a convenient way of normalizing patterns without the need to modify the data itself. 
offset 
The offset used in a polynomial kernel. 
algo 
the algorithm to solve the problem w.r.t. 
lambda 
regularization constant (default: 1). Rules the tradeoff
between regularization and correct classification on 
rho 
regularization constant (default: 1). 
eps 
epsilon in the termination condition. 
svmadmm
internally computing kernel matrix when type
is 0, which is based by the package kernlab.
An list containing the fitted model, including:
alpha 
A solution for dual form svm classification. 
beta 
A solution for primal form svm classifciation, also are the model weights. 
type 
An integer correspinding to 
kernel 
A function to define the kernel. 
x.tr 
The training input data. 
y.tr 
The training output data. 
1 2 3 4 5 6 7 8 9 10 11 12 13  library(svmadmm)
n = 100
p = 10
x = matrix(runif(2 * n * p, 1, 1), nrow = 2 * n)
y = sign(x[, 1])
y.ind = sample(1 : (2 * n), n / 10, replace = FALSE)
y[y.ind] =  y[y.ind]
x.tr = x[1 : n, ]
y.tr = y[1 : n]
x.te = x[(1 : n), ]
y.te = y[(1 : n)]
model = svm.admm(x.tr, y.tr)
fit = svm.predict(x.te, model)

Questions? Problems? Suggestions? Tweet to @rdrrHQ or email at ian@mutexlabs.com.
Please suggest features or report bugs with the GitHub issue tracker.
All documentation is copyright its authors; we didn't write any of that.