eLago {lago}R Documentation

Elliptical LAGO

Description

This is the original LAGO procedure, called “eLAGO” now to differentiate it from later variations such as “sLAGO”.

Usage

eLago(Y, X)
eLago(Y, X, D=preLago(Y,X), alpha = 0.5, K = 5, kernel="g")
eLago(Y, X, alpha = 0.5, K = 5, kernel="g") 

Arguments

Y a vector of binary responses; training data.
X a matrix of covariates/predictors; training data.
D if is.null(D)==T, the matrix D will be calculated by the function; however, it's computationally desirable to precompute D using preLago(Y, X), especially if the function eLago will be called repeatedly on the same data, e.g., during cross-validation.
alpha a positive real number; if alpha > 1, the radius/bandwidth is stretched; if alpha < 1, the radius/bandwidth is dampened.
K a positive integer indicating the number of nearest neighbors to use for calculating the radius/bandwidth.
kernel a character input; ‘t’ for “triangular”; ‘g’ for “Gaussian”; otherwise a “uniform” kernel is used but it important to note that a uniform kernel is not very effective.

Value

C1 an n1-by-p matrix whose rows are the centers of the LAGO radial basis function network.
R an n1-by-p matrix whose rows specify the radius vector of the n1 radial basis functions.
alpha the stretching/dampening parameter; see above; here passed on to be used for prediction.
kernel either ‘t’ or ‘g’; see above; here passed on to be used for prediction.

Author(s)

Alexandra Laflamme-Sanders and Mu Zhu, University of Waterloo, Canada.

References

Zhu M, Su W, Chipman HA (2006). LAGO: A computationally efficient approach for statistical detection. Technometrics, 48(2), 193 – 205.

Zhu M (2008). Kernels and ensembles: Perspectives on statistical learning. The American Statistician, 62(2), 97 – 109.

See Also

sLago, preLago, view.eLago, rank.eLago


[Package lago version 0.1-1 Index]