seibert.evolve: Evolve a population of parameter values
In paleolimbot/easyoptim: Easy Muti-Parameter Optimisation

Description Usage Arguments Value Examples

View source: R/gap_seibert.R

Evolve a population of parameter values

seibert.evolve(params, values, probs = NULL, rescaler = c("rescale_none",
  "rescale", "squish"), seed = NULL, validate = TRUE,
  breedargs = list(breed.use1 = 0.41, breed.use2 = 0.41, breed.rinterpolate =
  0.16, breed.mutate = 0.02))

`params`	A `list` of parameters with identical names to `values`
`values`	A `list` of parameter values with identical names to `params`. If all parameters are named parameters, this can be a data.frame.
`probs`	A vector of probabilities with which to weight sampling of the breeding couples. Can be NULL, indicating there is no preference.
`rescaler`	Pass TRUE to rescale `probs` to the range (0, 1). Or optionally a function that transorms probs to valid probability weights. This is useful if the value of probs used is sometimes negative from some objective function.
`seed`	Seed for random operations
`validate`	Pass FALSE to skip validation (may be faster).
`breedargs`	Passed to seibert.breed; may be overridden by parameters.

A list of parameters with identical names to params and values

params <- list(param.discrete(c('heads', 'tails')),
               param.real(c(0, 1)),
               beersdrank=param.normal.int(5, 1),
               taxes=param.fixed(TRUE))
values <- lapply(params, initial.value, n=10)

seibert.evolve(params, values)

# 'select' for heads
for(i in 1:10) {
  values <- seibert.evolve(params, values,
                           ifelse(values[[1]]=="heads", 0.6, 0.4))
}