LSVM: Linear Support Vector Machine under monotonicity constraints
In clemlaflemme/test: Methods in Structural Reliability

Description Usage Arguments Details Value Author(s) References See Also Examples

Produce a globally increasing binary classifier built from linear monotonic SVM

1	LSVM(x, A.model.lsvm, convexity)

`x`	a set of points where the class must be estimated.
`A.model.lsvm`	a matrix containing the parameters of all hyperplanes.
`convexity`	Either -1 if the set of data associated to the label "-1" is convex or +1 otherwise.

LSVM is a monotonic binary classifier built from linear SVM under the constraint that one of the two classes of data is convex.

An object of class integer representing the class of x

res

A vector of -1 or +1.

Vincent Moutoussamy

R.T. Rockafellar:
Convex analysis
Princeton university press, 2015.
N. Bousquet, T. Klein and V. Moutoussamy :
Approximation of limit state surfaces in monotonic Monte Carlo settings
Submitted .

modelLSVM

# A limit state function
f <- function(x){  sqrt(sum(x^2)) - sqrt(2)/2 }

# Creation of the data sets
n <- 200
X <- matrix(runif(2*n), nrow = n)
Y <- apply(X, MARGIN = 1, function(w){sign(f(w))})

#The convexity is known
## Not run: 
  model.A <- modelLSVM(X, Y, convexity = -1)
  m <- 10
  X.test <- matrix(runif(2*m), nrow = m)
  classOf.X.test <- LSVM(X.test, model.A, convexity = -1)

## End(Not run)