CODEWITHSUNDEEP
pythongenetic-algorithm
clement
clement

Reputation: 11

How to calculate population fitness using PyGAD?

I'd like to use PyGAD to solve job assignment problem and hope to get the result not only one chromosome but whole population. Such as [[1,1,0,0][0,0,1,1]] represent 4 slots,and 1 in first chromosome mean one guy assigned to those slots,another two slots assign to another guy according to second chromosome.

Follow the tutorial on the PyGAD website (https://pygad.readthedocs.io),am I doing right use initial_population to generate population？Is it right to use cal_pop_fitness() function ? I made some tries and still have no idea.

import pygad
import numpy

sol_per_pop = 2 
num_genes = 4 
pop_size = (sol_per_pop,num_genes) 

new_population = numpy.random.randint(low=0,high=2,size=pop_size) 

num_generations = 100
num_parents_mating = 2

def fitness_func(solution,solution_idx):
    fitness = ga_instance.cal_pop_fitness(new_population)
    #do something
    

ga_instance = pygad.GA(initial_population = new_population,
                       num_generations=num_generations,
                       num_parents_mating=num_parents_mating,
                       sol_per_pop=sol_per_pop,
                       num_genes=num_genes,
                       mutation_type=None,
                       fitness_func=fitness_func)
                       
ga_instance.run()

Upvotes: 1

Views: 1282

Answers (1)

Ahmed Gad
Ahmed Gad

Reputation: 866

I have some notes:

  1. cal_pop_fitness() is not a function. It is a method in the pygad.GA class.
  2. The cal_pop_fitness() method does not accept any parameters. It is not correct to pass the new_population to it.
  3. This is a critical note. The fitness function you pass to the fitness_func parameter calls the cal_pop_fitness() method. In PyGAD, the cal_pop_fitness() method calls your custom fitness function. Thus, you enter an infinite loop where the 2 functions are calling each other.

So, your code cannot work this way.

If you want to get the fitness of the most recent population, you can use the last_generation_fitness attribute. Just access it within your fitness function:

ga_instance.last_generation_fitness

Note that this attribute is None for the first generation. So, you can set it to some initial fitness values before calling the run() method:

ga_instance.last_generation_fitness = numpy.array([1, 1])

Upvotes: 2

Related Questions