Skip to first unread message
tom p.
Aug 13, 2018, 6:25:51 AM8/13/18
to deap-users
I have a premature convergence problem. I want to increase the diversity of my population. My individuals consist of a list of 20 numbers between 0 and 3. My parameters are:
toolbox.register("evaluate", maxFitness)
toolbox.register("mate", tools.cxTwoPoint)
toolbox.register("mutate",tools.mutUniformInt, low=0, up=3, indpb=0.05)
toolbox.register("select", tools.selNSGA2)
NGEN = 250
MU = 200
LAMBDA = 400
CXPB = 0.8
MUTPB = 0.2
pop = toolbox.population(n=MU)
toolbox.register("evaluate", maxFitness)
toolbox.register("mate", tools.cxTwoPoint)
toolbox.register("mutate",tools.mutUniformInt, low=0, up=3, indpb=0.05)
toolbox.register("select", tools.selNSGA2)
NGEN = 250
MU = 200
LAMBDA = 400
CXPB = 0.8
MUTPB = 0.2
pop = toolbox.population(n=MU)
I want to increase the diversity of my population to avoid early convergence. I saw cxSimulatedBinaryBounded has an eta value, which can be modified to produce children that are different to their parents. The documentation says it expects sequence individuals of floating point numbers, why is this the case? Would it not work for whole numbers?
Cheers.
Reply all
Reply to author
Forward
0 new messages