fit.svm: fit.svm

In kalinn/weightedSVM: Weighted Support Vector Machines

Description

Fit support vector machine with optional subject-level weights

Usage

fit.svm(tr.x, tr.y, ts.x, grid.C, w = NULL)

Arguments

tr.x	Matrix or data.frame of features for training
tr.y	Vector of group labels for training
ts.x	Matrix or data.frame of features for testing
grid.C	Vector of C values for tuning the SVM
w	optional vector of subject weights

Author(s)

Kristin Linn

Examples

if (require(MASS)){
##################################################
# Generate data
##################################################
set.seed(1)
n = 500
p = 2
sigma = matrix(.25, p, p)
diag(sigma) = 1
eps1 = MASS::mvrnorm(n, rep(0, p), sigma)
eps2 = MASS::mvrnorm(n, rep(0, p), sigma)
tr.y = rbinom(n, 1, .5)
x1 = 5 - .25*tr.y + .2*tr.y*eps1[,1] + .1*eps2[,1]
x2 = 5 - tr.y + .3*eps2[,1] + .5*(eps1[,2] + eps2[,2])
tr.x = cbind(x1, x2)
ts.eps1 = MASS::mvrnorm(n, rep(0, p), sigma)
ts.eps2 = MASS::mvrnorm(n, rep(0, p), sigma)
ts.y = rbinom(n, 1, .5)
ts.x1 = 5 - .25*ts.y + .2*ts.y*ts.eps1[,1] + .1*ts.eps2[,1]
ts.x2 = 5 - ts.y + .3*ts.eps2[,1] + .5*(ts.eps1[,2] + ts.eps2[,2])
ts.x = cbind(ts.x1, ts.x2)
grid.C = 10^c(-3:3)
out = fit.svm(tr.x=tr.x, tr.y=tr.y, ts.x=ts.x, grid.C=grid.C)
w = c(rep(2, 250), rep(1.5, 125), rep(3, 125))
w.out = fit.svm(tr.x=tr.x, tr.y=tr.y, ts.x=ts.x, grid.C=grid.C, w=w)
}

kalinn/weightedSVM documentation built on May 20, 2019, 6:33 a.m.