gptk (version 1.06)

expTransform: Constrains a parameter.

Description

contains commands to constrain parameters to be positive via exponentiation or within a fixed interval via the sigmoid function.

Usage

expTransform(x, transform)
sigmoidTransform(x, transform)
boundedTransform(x, transform, bounds)

Arguments

x

input argument.

transform

type of transform, 'atox' maps a value into the transformed space (i.e. makes it positive). 'xtoa' maps the parameter back from transformed space to the original space. 'gradfact' gives the factor needed to correct gradients with respect to the tran

bounds

a 2-vector of bounds of allowed values in boundedTransform

Value

• Return value as selected by tranform

See Also

modelOptimise

Examples

# Transform unconstrained parameter -4 to a positive value
  expTransform(-4, 'atox')

  # Transform a bounded parameter in (1,3) to an unconstrained one
  boundedTransform(2, 'xtoa', c(1, 3))