plgp (version 1.0)

exp2d.C: 2-d Exponential Hessian Data

Description

Generates 2-d classification data with two or three class labels, based on the Hessian data from a 2-d real-valued response

Usage

exp2d.C(X, threed = TRUE)

Arguments

X

a matrix or data.frame describing the design at which the response categories are desired

threed

a scalar logical indicating if the two or three-class version of the class labels should be returned.

Value

• A vector of class labels of length nrow(X) is returned

Details

The underlying real-valued response is governed by $$Z(X)=x_1 * \exp(x_1^2-x_2^2).$$ Two class labels are generated by inspecting the sign of the sum of the eigenvalues of the Hessian (Broderick & Gramacy, 2010). This generates the first (-) and second (+) classes in a three-class function. A third class label (the default) may created from the first one where X[,1] > 0 (Gramacy & Polson, 2010)

References

Broderick, T. and Gramacy, R. (2010). Classification and categorical inputs with treed Gaussian process models. Tech. rep., University of Cambridge. ArXiv:0904.4891. Gramacy, R. and Polson, N. (2010). Particle learning of Gaussian process models for sequential design and optimization. Tech. Rep. arXiv:0909.5262, University of Cambridge. http://www.statslab.cam.ac.uk/~bobby/plgp.html

Examples

## The following demos use this data
## Illustrates classification GPs on a simple 2-d exponential
## data generating mechanism
demo("plcgp_exp", ask=FALSE)

## Illustrates active learning via entropy with classification
## GPs on a simple 2-d exponential data generating mechanism
demo("plcgp_exp_entropy", ask=FALSE)