lineqGPSys: Linear Systems of Inequalities for '"lineqGP"' Models

In lineqGPR: Gaussian Process Regression Models with Linear Inequality Constraints

Description

Build the linear system of inequalities for "lineqGP" models.

Usage

lineqGPSys(
  m = nrow(A),
  constrType = c("boundedness", "monotonicity", "convexity", "linear"),
  l = -Inf,
  u = Inf,
  A = diag(m),
  d = length(m),
  lineqSysType = "twosides",
  constrIdx = seq(length(m)),
  rmInf = TRUE
)

Arguments

m	the number of linear inequality constraints.
constrType	a character string corresponding to the type of the inequality constraint. Options: "boundedness", "monotonicity", "convexity", "linear"
l	the value (or vector) with the lower bound.
u	the value (or vector) with the upper bound.
A	a matrix containing the structure of the linear equations.
d	the value with the input dimension.
lineqSysType	a character string corresponding to the type of the linear system. Options: twosides, oneside (see bounds2lineqSys for more details).
constrIdx	for d > 1, a logical vector with the indices of active constrained dimensions.
rmInf	If TRUE, inactive constraints are removed (e.g. -Inf ≤ x ≤ Inf).

Value

A list with the linear system of inequalities: list(A,l,u) (twosides) or list(M,g) (oneside).

Comments

This function could change in future versions for more types of inequality constraints in higher dimensions.

Author(s)

A. F. Lopez-Lopera.

References

Lopez-Lopera, A. F., Bachoc, F., Durrande, N., and Roustant, O. (2017), "Finite-dimensional Gaussian approximation with linear inequality constraints". ArXiv e-prints [link]

See Also

bounds2lineqSys

Examples

linSys1 <- lineqGPSys(m = 5, constrType = "boundedness", l = 0, u = 1, lineqSysType = "twosides")
linSys1
linSys2 <- lineqGPSys(m = 5, constrType = "boundedness", l = 0, u = 1, lineqSysType = "oneside")
linSys2

lineqGPR documentation built on Jan. 11, 2020, 9:23 a.m.