# Copyright 2006 by Sean Luke and George Mason University # Licensed under the Academic Free License version 3.0 # See the file "LICENSE" for more information parent.0 = ../simple/simple.params # # The following parameter file will set up (mu,lambda) evolution, # single-threaded, no coevolution, no exchanging, # no cross-population breeding, using generational evolution, # simple fitness, popsize=1024, etc. # You'll need to fill in some of the gaps. # # This file tweaks some features in simple.params. Perhaps you might # change parent.0 to ../gp/koza/koza.params if you're doing GP (see below) # # Some notes: # 1. To perform truncation selection, include the ec.es.ESSelection # selection method in your breeding pipeline. You can mix in other # selection methods along with it. # # 2. You need to specify the mu and lambda values for each of your # subpopulations. This is different than past versions of ECJ where # you didn't specify the lambda -- it was inferred from the subpopulation # size. However, now, you are permitted to have whatever initial # subpopulation size you like. # # 3. If you want to do EP-style 50/50 truncation selection, I believe the # way to do it is to set your mu value to the same as the lambda, # and use ec.es.MuPlusLambdaBreeder. # # We begin by specifying the mu value for the first subpopulation, # You need to change this and/or specify mu values for additional # subpopulations as needed. We'll default the mu to 1. # We'll default the lambda to 10. What the heck. # ================================================================ es.mu.0 = 1 es.lambda.0 = 10 breed = ec.es.MuCommaLambdaBreeder # you'll need to remember to set up the system so that one, and exactly # one, selection method called each time is an ESSelectionMethod. Here's # how you could do it if you were doing Koza-style GP stuff: # gp.koza.reproduce.source.0 = ec.es.ESSelection # gp.koza.xover.source.0 = ec.es.ESSelection # gp.koza.xover.source.1 = ec.select.TournamentSelection # gp.koza.mutate.source.0 = ec.es.ESSelection