python create genetic algorithm no convergence to maximum or minimum

Question

Hello i need help for create genetic algorithme for converge to maximum or minimum value. I develop a code for found maximum sentence ascii sum, but my code not converge to maximum, my code make "yoyo" value

like this picture : matploltib output

i share my code :

import random
import statistics

EVOLUTION=[]

words = [
        ["Un", "Des", "Une", "On", "Elle"],
        ["a", "eu", "avait", "est", "était", "fut"],
        ["soif", "rouge"]
        ]

def individual(data):
    #return tuple(random.choice(range(len(feature))) for feature in data)
    return tuple(random.choice(range(len(feature))) for feature in data)


def population(data, initial=100):
    return [individual(data) for i in range(initial)]


def fitness(individual, data):
    chaine=sentence(individual,words)
    somme = 0
    for caractere in chaine:
        somme = somme + ord(caractere)

    print(chaine)
    print(somme)
    EVOLUTION.append(somme)
    return somme
    #return sum(data[i][individual[i]] for i in range(len(individual)))


def grade(population, data):
    fit = [fitness(ind, data) for ind in population]
    return statistics.mean(fit)


def mutate(ind, data):
    gene = random.randrange(0, len(ind))
    clone = list(ind)
    clone[gene] = random.randrange(0, len(data[gene]))
    #print(sentence(tuple(clone),words))
    return tuple(clone)


def cross(mother, father):
    return tuple(round(statistics.mean(genes)) for genes in zip(mother, father))

def sentence(individual, words):
    return ' '.join([words[i][individual[i]] for i in range(len(words))])

def evolve(population, data, retain=0.0, random_select=0.00, mutation_rate=0.00):
    def cmp_ind(ind):
        return fitness(ind, data)
    sorted_population = sorted(population, key=cmp_ind, reverse=True)

    len_retained = round(len(population) * retain)
    retained = sorted_population[:len_retained]

    random_selected = [
        ind
        for ind in sorted_population[len_retained:]
        if random.random() <= random_select
    ]

    mutated = [
        mutate(ind, data)
        for ind in sorted_population[len_retained:]
        if random.random() <= mutation_rate
    ]

    children = [
        cross(random.choice(sorted_population),
              random.choice(sorted_population))
        for i in range(len(population) - len(random_selected) - len(mutated))
    ]

    return random_selected + mutated + children




if __name__ == '__main__':

    data = [[len(w) for w in ws] for ws in words]



    initial_population = population(data, 30)
    next_population = initial_population
    max_iter = 3

    for i in range(max_iter):
        next_population = evolve(next_population, data)

    sorted_population = sorted(next_population, key=lambda x: fitness(x, data))
    best_individual = sorted_population[0]

    print("best solution :")

    chaine=sentence(best_individual,words)
    somme = 0
    for caractere in chaine:
        somme = somme + ord(caractere)
    print(chaine)
    print(somme)

    import matplotlib
    matplotlib.use('Agg')
    import matplotlib.pyplot as plt
    plt.plot(EVOLUTION)
    plt.savefig('myfig')

i want to found highter solution in my fitness function

thanks for advance for your help

Welcome to StackOverflow. Please read and follow the posting guidelines in the help documentation, as suggested when you created this account. [Minimal, complete, verifiable example](http://stackoverflow.com/help/mcve) applies here. We cannot effectively help you until you post your MCVE code and accurately describe the problem. We should be able to paste your posted code into a text file and reproduce the problem you described. — Prune, Apr 04 '18 at 14:53
Specifically, you gave us the entire program, rather than a minimal example. Second, you describe your output as "yoyo", but "yoyo" does not appear anywhere in your output. Third, your output is in French; this is an English-language site, and only a few of us read even the simple French you use here. Finally, we need you to minimize the problem: give us the shortest code necessary to reproduce it. If your problem is in the fitness function, then unit-test that function, and leave all of the genetic parts out of the posting. — Prune, Apr 04 '18 at 14:55
@Prune I think „yoyo“ is used to describe the progressing of the value, like the movement of the [yo-yo](https://en.wikipedia.org/wiki/Yo-yo) toy. It's not converging but bouncing up and down. — BlackJack, Apr 04 '18 at 16:50
@BlackJack: that's what I hoped when I read it, but there's no specific output in the posting to clarify. Even though I read French well enough to follow the text, I'm hopeful that a "proper" example will get the problem solved quickly. The photo shows the effect, but there's nothing about what the function *should* return. In my DL work, there's generally nothing wrong with that graph for the first 120 iterations. — Prune, Apr 04 '18 at 17:09

python create genetic algorithm no convergence to maximum or minimum

0 Answers0