Maximization isted of minimization

[GrzegorzGiniewicz]

Hi.

i have problem with minimization of ackley benchmark function. I cant achieve minimization.

im using

creator.create("FitnessMin", base.Fitness, weights=(-1.0,))
creator.create("Individual", list, typecode='d', fitness=creator.FitnessMin)

Im loking for help on stackoveflow. There is my code. Link:

https://stackoverflow.com/questions/44576799/minimize-a-function-using-deap

[fmder]

You must use axis=0 when registering numpy stats operators. It should get you a better insight. Otherwise, your code look fine to me and there is no bug on that aspect in DEAP.

[GrzegorzGiniewicz]

I used axis = 0 but it doesn't help. I need to use SelRoulette for my program but it still going for maximization, no matter if i put in weights 1.0 or -1.0 or use FitnessMin or FitnessMax. When i use SelTournament, tournsize = 3 i can change weights and it's maximization for 1.0 and minimization for -1.0. Why it doesnt work with SelRoulette ? What can i do to achive minimization and use SelRoulette? @fmder

[fmder]

Ho yeah, roulette selection cannot be used for minimizing has it cannot handle negative numbers. Le 17 juin 2017 07:07, "GrzegorzGiniewicz" <notifications@github.com> a écrit :

…

I used axis = 0 but it doesn't help. I need to use SelRoulette for my program but it still going for maximization, no matter if i put in weights 1.0 or -1.0 or use FitnessMin or FitnessMax. When i use SelTournament, tournsize = 3 i can change weights and it's maximization for 1.0 and minimization for -1.0. Why it doesnt work with SelRoulette ? What can i do to achive minimization and use SelRoulette? — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#211 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AA6rwsjyszBQbETVoF5VVi0qphqACa8Iks5sE7NVgaJpZM4N8SpT> .

[GrzegorzGiniewicz]

Thanks for quick answer and sorry for trouble.

[fmder]

Maybe you can try #87, if it works you could propose a pull request.

[GrzegorzGiniewicz]

I've changed selection for SelBest but i cant set k= 20 [like this: (toolbox.register("select", tools.selBest, k=20) ] im geting error: "selBest" got multiple values for keyword argument 'k'. And still its maximizing, even with :
creator.create("FitnessMin", base.Fitness, weights=(-1.0,))
creator.create("Individual", list, typecode='d', fitness=creator.FitnessMin)

I understand that k is for the number of individuals to select.
Without k parametr in selection it works but still maximizing.

import random
import numpy
from deap import algorithms
from deap import base
from deap import creator
from deap import tools
from deap import benchmarks
from random import Random
from time import time
import math

toolbox = base.Toolbox()  
ind = 5
IND_SIZE = ind

creator.create("FitnessMin", base.Fitness, weights=(-1.0,))
creator.create("Individual", list, typecode='d', fitness=creator.FitnessMin)

toolbox.register("expr", random.sample, range(IND_SIZE), IND_SIZE,)

toolbox.register("individual", tools.initIterate, creator.Individual, toolbox.expr)
toolbox.register("population", tools.initRepeat, list, toolbox.individual)

toolbox.register("mate", tools.cxTwoPoint)
toolbox.register("mutate", tools.mutGaussian, mu=0.2, sigma=0.2, indpb=0.1)
toolbox.register("select", tools.selBest)  #k parameter doesnt work
toolbox.register("evaluate", benchmarks.ackley)


def main():
prng=None
best=None

if prng is None:
    prng = Random()
    prng.seed(time())

random.seed(prng)
pop = toolbox.population(n=100)

stats = tools.Statistics(lambda ind: ind.fitness.values)
stats.register("best", numpy.mean, axis=0)

algorithms.eaSimple(pop, toolbox, 0.9, 0.2, 100, stats=stats)

return pop, best

pop,best= main()

[fmder]

In eaSimple, selection is made to replace the whole population with k=len(population).

[fmder]

If you want to customize selection with multiple operators I'd say you should copy the content of eaSimple in your own script and replace the selection part, you'll be able to get exactly what you want really easily. The content of eaSimple is no more than 30 lines of code.

[GrzegorzGiniewicz]

ok i changed k for 20 but still it's maximizing. How i can achive minimization ? @fmder

import random
import numpy
import operator
from deap import algorithms
from deap import base
from deap import creator
from deap import tools
from deap import gp
from deap import benchmarks
from random import Random
from time import time
import math
 
 
def varAnd(population, toolbox, cxpb, mutpb):
 
 
    offspring = [toolbox.clone(ind) for ind in population]
# Apply crossover and mutation on the offspring
    for i in range(1, len(offspring), 2):
        if random.random() < cxpb:
            offspring[i - 1], offspring[i] = toolbox.mate(offspring[i - 1], offspring[i])
            del offspring[i - 1].fitness.values, offspring[i].fitness.values
 
    for i in range(len(offspring)):
        if random.random() < mutpb:
            offspring[i], = toolbox.mutate(offspring[i])
            del offspring[i].fitness.values
 
    return offspring
 
 
def eaSimple(population, toolbox, cxpb, mutpb, ngen, stats=None, halloffame=None, verbose=__debug__):
 
    logbook = tools.Logbook()
    logbook.header = ['gen', 'nevals'] + (stats.fields if stats else [])
 
# Evaluate the individuals with an invalid fitness
    invalid_ind = [ind for ind in population if not ind.fitness.valid]
    fitnesses = toolbox.map(toolbox.evaluate, invalid_ind)
    for ind, fit in zip(invalid_ind, fitnesses):
        ind.fitness.values = fit
 
    if halloffame is not None:
        halloffame.update(population)
 
    record = stats.compile(population) if stats else {}
    logbook.record(gen=0, nevals=len(invalid_ind), **record)
    if verbose:
        print logbook.stream
 
# Begin the generational process
    for gen in range(1, ngen + 1):
# Select the next generation individuals
        offspring = tools.selBest(population, k=20)
# Vary the pool of individuals
        offspring = varAnd(offspring, toolbox, cxpb, mutpb)
# Evaluate the individuals with an invalid fitness
        invalid_ind = [ind for ind in offspring if not ind.fitness.valid]
        fitnesses = toolbox.map(toolbox.evaluate, invalid_ind)
        for ind, fit in zip(invalid_ind, fitnesses):
            ind.fitness.values = fit
# Update the hall of fame with the generated individuals
        if halloffame is not None:
            halloffame.update(offspring)
# Replace the current population by the offspring
        population[:] = offspring
# Append the current generation statistics to the logbook
        record = stats.compile(population) if stats else {}
        logbook.record(gen=gen, nevals=len(invalid_ind), **record)
        if verbose:
            print logbook.stream
 
    return population, logbook
 
 
 
def main():
 
    toolbox = base.Toolbox()  
    ind = 5
    IND_SIZE = ind
 
    creator.create("FitnessMin", base.Fitness, weights=(-1.0,))
    creator.create("Individual", list, typecode='d', fitness=creator.FitnessMin)
 
    toolbox.register("expr", random.sample, range(IND_SIZE), IND_SIZE)
 
    toolbox.register("individual", tools.initIterate, creator.Individual, toolbox.expr)
    toolbox.register("population", tools.initRepeat, list, toolbox.individual)
 
    toolbox.register("mate", tools.cxTwoPoint)
    toolbox.register("mutate", tools.mutGaussian, mu=0.2, sigma=0.2, indpb=0.1)
    #toolbox.register("select", tools.selBest)
    toolbox.register("evaluate", benchmarks.ackley)
 
 
 
    prng=None
    best=None
 
    if prng is None:
        prng = Random()
        prng.seed(time())
       
    random.seed(prng)
    pop = toolbox.population(n=100)
 
    stats = tools.Statistics(lambda ind: ind.fitness.values)
    stats.register("best", numpy.mean, axis=0)
 
    eaSimple(pop, toolbox, 0.9, 0.2, 100, stats=stats)
 
#verbose=False
 
    return pop, best
 
pop,best= main()

[fmder]

How do you see that it is maximizing?

[GrzegorzGiniewicz]

im printing results :
https://pastebin.com/XmSeegsV
best sholud be coming to 0

Its wird that its not working, when i use selTournament it's fine . Buy i rly need selBest

[fmder]

I just tested your script. I just saw that the "best" you are printing is the average of all fitnesses. What you really should look for is the min value. Add those lines to your script

stats.register("min", numpy.min, axis=0)
stats.register("max", numpy.max, axis=0)
stats.register("avg", numpy.average, axis=0)

You'll see that the min value starts around 7 or so and goes to 0 in about 5 to 10 generations. Why does it go up again? It is because of the And variation operator and your high crossover and mutation rate. The evaluation is done after variation and before selection thus you might modify the best individuals. However by using a hall of fame you will still capture it.

With high crossover rate it is possible for the max value to go up and the min value to go down averaging the optimisation result over the entire population, exactly what you are observing. Essentially you should just check for the min value.

[GrzegorzGiniewicz]

Thank you very much for help and patiance.

[fmder]

:) 2017-06-17 10:07 GMT-04:00 GrzegorzGiniewicz <notifications@github.com>:

…

Thank you very much for help and patiance. — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#211 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AA6rwqe-pw-RRN0ghodUB_HtUhHqYOmFks5sE929gaJpZM4N8SpT> .