make_runner_function: Factory for objective functions.
In jakobbossek/evoprob: Evolutionary Algorithm to Produce Diverse Problems

Description Usage Arguments Value

View source: R/make_runner_function.R

Given a set of n algorithms A (parameter A) with parameters θ_i, i \in A (parameter A.pars) the goal is to produce a function f = function(x, ...) that expects a problem instance x and optional arguments .... The function shall run each algorithm A_i \in A parameterized with θ_i on x. In case of stochastic/randomized algorithms each algorithm should be run multiple times (parameter repls) returning an aggregated value (parameter aggr.fun).

make_runner_function(
  A,
  A.pars,
  runner.fun,
  repls = 1L,
  aggr.fun = base::mean,
  named.result = TRUE
)

`A`	[`character`] Character vector of at least two algorithm names. Note that the names must be valid arguments for the “algorithm” parameter of parameter `runner.fun`.
`A.pars`	[`list of lists`] A named list (names equal to `A`) where each sub-list should contain parameters for the respective algorithm.
`runner.fun`	[`function(x, algorithm, ...)`] A function that expects an instance `x`, an algorithm name `algorithm` (see parameter `A`) and optional further parameters (see parameter `A.pars`).
`repls`	[`integer`] How many independent runs should be conducted for each algorithm on an instance? Defaults to 1, but should be adjusted if the corresponding algorithm(s) is/are of stochastic nature. If a single value is passed it is used for all algorithms.
`aggr.fun`	[`function(x, ...)`] Function used to aggregate results of multiple runs. Only applied if `repls[i]` is greater than one for algorithm `A[i]`. Defaults to `mean`.
`named.result`	[`logical(1)`] Should the result vector be named with `A`? Default is `TRUE`.