lasvmR (version 0.1.2)

lasvmTrain: lasvmTrain

Description

Use lasvm to train a given problem.

Usage

lasvmTrain(x, y, gamma = 1, cost = 1, degree = 3, coef0 = 0, optimizer = 1, kernel = 2, selection = 0, termination = 0, sample = 1e+08, cachesize = 256, bias = 1, epochs = 1, epsilon = 0.001, verbose = FALSE)

Arguments

x

data matrix

y

labels

gamma

RBF kernel parameter

cost

regularization parameter

degree

degree for poly kernel

coef0

coefficient for poly kernel

optimizer

type of optimizer

kernel

kernel type

selection

selection strategy

termination

criterion for stopping

sample

time for stopping/number of iterations tec

cachesize

size of kernel cache

bias

use bias?

epochs

number of epochs

epsilon

stopping criterion parameter

verbose

verbose output?

Value

a list consisting of alpha alpha for SVs as vector SV support vectors as matrix

Examples

model = lasvmR::lasvmTrain (x = as.matrix(iris[seq(1,150,2),1:4]),
	y = (as.numeric(iris[seq(1,150,2),5]) %% 2)*2-1,
	gamma = 1,
	cost = 1,
	kernel = 2)
ytrue = (as.numeric(iris[seq(2,150,2),5]) %% 2)*2-1
result = lasvmPredict (x = as.matrix(iris[seq(2,150,2),1:4]), model)
ypred = result$predictions
error = sum(abs(ypred - ytrue))/length(ytrue)
cat ("Error rate =", error*100)