Skip to content

Setting Up GA

Jump to bottom
Anthony Awuley edited this page Jun 26, 2016 · 2 revisions

Start here for GA

The parameter files are located in the directory io/params/ and the main class is in ec.main.Run. Three tutorials have been set up and can be found in the directories

Setting up GA TSP

GENERAL parameters

  • number-of-experiments = 2
  • generations = 1000
  • population-size = 200
  • stop-when-solved = true
  • crossover-probability = 0.95
  • mutation-probability = 0.05
  • tournament-selection-pressure = 0.9
  • seed = 20
  • stop-when-solved = true

SELECTION Operator

  • selection-operation = ec.operator.operations.selection.TournamentSelection
  • tournament-size = 3

REPRODUCTION operators

Probability of crossover. State if the point of crossover is fixed on both chromosomes (The size will always be the same) or not

  • mutation_operation = ec.operator.operations.IntFlipMutation
  • elite-size = 2

The chromosomes are fixed length structures and doesnt change to mutation and crossover. The initialiser will random create integer strings in this case of length specified

  • initial-chromosome-size = 280

FITNESS function - Fitness function requires the full package name fitness-function = ec.fitnessevaluation.eval.TSP

EC Main class. This directs the system to the main evolving class. In this case the GA evolution

  • main-class = ec.algorithms.ga.Evolve

INITIALISATION - Specifies which module to use as an initialiser initial population.

  • initialiser = ec.operator.initialiser.Initialise
  • gene-representation = ec.individuals.representation.VRPTW deprecated

REPLACEMENT

  • replacement-operation = ec.operator.operations.replacement.Generational

CROSSOVER MODULE

  • crossover-operation = ec.operator.crossover.CycleCrossover
  • #or
  • crossover-operation = ec.operator.crossover.UniformOrderCrossover

MUTATION MODULE

  • mutation-operation = ec.operator.mutation.Inversion well tested
  • #or
  • mutation-operation = ec.operator.mutation.ReciprocalExchange

Specification of output statistics files

  • number-of-individuals = 2
  • stats-operation = ec.util.statistics.singleobjective.TSPStatistics

Nodes Information

  • nodes = 280
  • start-node = 1
  • #statistics prefix
  • stat-out = a280

Data File

  • #NAME : a280
  • #COMMENT : drilling problem (Ludwig)
  • #TYPE : TSP
  • #DIMENSION: 280
  • #EDGE WEIGHT TYPE : EUC 2D
  • 1 288 149
  • 2 288 129
  • 3 270 133
  • 4 256 141
  • 280 280 133

Evolutionary Computation

Clone this wiki locally