selection.py

# Create a random population of genes, with each genotype having 2 elements.
population_size = 200
population = np.random.standard_normal((population_size, 2))

# Some objective measure of the genes performance or fitness. This is dependant
# on the environment or optimsation problem that you have, and I do not define
# the function in this snippet.
fitnesses = evaluate_fitness(population)

# Sort the list of fitnesses and create a probability distribution based on
# how fit each gene is.
fitnesses_indices = fitnesses.argsort()
sorted_fitnesses = fitnesses[fitnesses_indices]
fitnesses_weighting = np.maximum(0, 1 - sorted_fitnesses / self.fitnesses.sum())
fitnesses_weighting /= fitnesses_weighting.sum()

# Sort the population by their fitness.
sorted_population = population[fitnesses_indices]

# How many new genes to make and a container to hold them.
amount_of_new_genes = 100
new_population = []

# Loop and select successful parents.
for _ in range(amount_new):

    # The probability of selecting a parent is proprotonate to the parents
    # objective fitness.
    i0 = np.random.choice(sorted_population.shape[0], p=fitnesses_weighting)
    i1 = np.random.choice(sorted_population.shape[0], p=fitnesses_weighting)
    
    # We now have parents, ready to 'mate'
    parent_one = population[i0]
    parent_two = population[i1]