makeAgeFitnessComplexityParetoGpSearchHeuristic: Age Fitness Complexity Pareto GP Search Heuristic for RGP

Description Usage Arguments Value

View source: R/search_heuristics.r

Description

The search-heuristic, i.e. the concrete GP search algorithm, is a modular component of RGP. makeAgeFitnessComplexityParetoGpSearchHeuristic creates a RGP search-heuristic that implements a generational evolutionary multi objective optimization algorithm (EMOA) that selects on three criteria: Individual age, individual fitness, and individual complexity.

Usage

1
2
3
4
5
6
7
makeAgeFitnessComplexityParetoGpSearchHeuristic(lambda = 50,
  crossoverProbability = 0.5, enableComplexityCriterion = TRUE,
  enableAgeCriterion = FALSE, ndsParentSelectionProbability = 0,
  ndsSelectionFunction = nds_cd_selection, complexityMeasure = function(ind,
  fitness) fastFuncVisitationLength(ind), ageMergeFunction = max,
  newIndividualsPerGeneration = if (enableAgeCriterion) 50 else 0,
  newIndividualsMaxDepth = 8, newIndividualFactory = makePopulation)

Arguments

lambda

The number of children to create in each generation (50 by default).

crossoverProbability

The crossover probability for search-heuristics that support this setting (i.e. TinyGP). Defaults to 0.5.

enableComplexityCriterion

Whether to enable the complexity criterion in multi-criterial search heuristics.

enableAgeCriterion

Whether to enable the age criterion in multi-criterial search heuristics.

ndsParentSelectionProbability

The probability to use non-dominated sorting to select parents for each generation. When set to 0.0, parents are selected by uniform random sampling without replacement every time. Defaults to 1.0.

ndsSelectionFunction

The function to use for non-dominated sorting in Pareto GP selection. Defaults to nds_cd_selection.

complexityMeasure

The complexity measure, a function of signature function(ind, fitness) returning a single numeric value.

ageMergeFunction

The function used for merging ages of crossover children, defaults to max.

newIndividualsPerGeneration

The number of new individuals per generation to insert into the population. Defaults to 50 if enableAgeCriterion == TRUE else to 0.

newIndividualsMaxDepth

The maximum depth of new individuals inserted into the population.

newIndividualFactory

The factory function for creating new individuals. Defaults to makePopulation.

Value

An RGP search heuristic.


rgp documentation built on May 30, 2017, 12:45 a.m.