## Introduction

lasvmR is a simple wrapper for the LASVM Solver (see [http://leon.bottou.org/projects/lasvm]). LASVM is basically an online variant of the SMO solver, but citing the original webpage is better:

LASVM is an approximate SVM solver that uses online approximation. It reaches accuracies similar to that of a real SVM after performing a single sequential pass through the training examples. Further benefits can be achieved using selective sampling techniques to choose which example should be considered next.

Note that LASVM works with binary labels only, Y = {1, -1}. Nearly all parameters have been carried over. Multi-model saving was disabled.

## Example

library(lasvmR)

# generate two gaussian clusters (very easy example)
qx = rnorm(100, mean = -3, sd = 1) - 1
qy = rnorm(100, mean = -3, sd = 1) - 1
px = rnorm(100, mean = 3, sd = 1) + 1
py = rnorm(100, mean = 3, sd = 1) + 1
data = rbind( cbind(px, py), cbind(qx, qy) )
label = sign (data[,1])
model = lasvmTrain (x = data, y = label, gamma = 1.0, cost = 1.0, epochs = 33, epsilon = 0.01)
result = lasvmPredict (data, model)

# compute error
error = sum(abs(label - result\$predictions))/length(label)*100
cat ("Error rate is ", error)

Error rate is 0