The x matrix for the training dataset.
Columns represent the covariates, and rows represent
the instances. There should be no NA/NaN values in x.
y
The labels for the training dataset.
gamma
The convex combination parameter of the loss function.
weight
The weight vector for each observation. By default,
the program uses equal weights for all observations.
lambda
The user specified lambda values.
kernel
The kernel for classification.
kparam
The kernel parameter. If kernel=linear, this option
is ignored. For kernel=polynomial, it is the order
of the polynomial functions. For kernel=gaussian,
it is the Gaussian kernel parameter.
large
Whether the number of observations is large in the
data. If TRUE, then the algorithm will split the
data set into several parts and train on each part
to provide a warm start for the entire data training.
This option aims to enhance the computational speed.
epsilon
Convergence threshold in coordinate descent circling
algorithm. The smaller epsilon is, the more accurate
the final model is, and the more time it takes for
calculation. Default is
(0.0001*number of observations*number of classes).
warm
A matrix that contains the warm start for slack
variables alpha. This option is especially useful
when the user wishes to obtain the classifier with
higher level accuracy (smaller epsilon) or with a
different lambda, if the warm start is available
from an existing ramsvm output.
nb.core
The number of threads to use for parallel computing.
If null, the code will automatically detect and use
the number of CPU cores. This option is used only
when large=TRUE.
Value
An object of class ramsvm is returned.
If kernel=linear, this S4 object contains the following:
x
A copy of the input covariate matrix.
y
A copy of the input labels.
y.name
The class names of y.
k
Number of classes in the classification problems.
gamma
A copy of the convex combination parameter of the loss
function.
weight
The weight vector for each observation.
lambda
The lambda vector of all lambdas in the solution path.
beta
A list of matrices containing the estimated parameters
of the classification function. Each matrix in the
list corresponds to the lambda value in the solution
path in order. For one single matrix, the rows
correspond to a specific predictor.
beta0
A list of the intercepts of the classification
function. Each vector in the list corresponds to
the lambda in the solution path in order.
epsilon
Convergence threshold in coordinate descent circling
algorithm.
call
The call of ramsvm.
If kernel != linear the S4 object also contains the following:
kernel
The kernel for classification.
kparam
The kernel parameter.
Author(s)
Chong Zhang, Yufeng Liu, and Shannon Holloway
References
C. Zhang, Y. Liu, J. Wang and H. Zhu. (2015+).
Reinforced Angle-based Multicategory Support Vector Machines.
Journal of Computational and Graphical Statistics, in press.