CODEWITHSUNDEEP
pythongenetic-algorithm
eignvalue
eignvalue

Reputation: 23

Passing arguments to PyGAD fitness function

I'm curious how you send arguments to the fitness function in PyGAD i.e

import pygad

def fitness_function(solution, solution_idx,num):
    print(num+10)
    return sum(solution)

ga_instance = pygad.GA(num_generations=1,
                       num_parents_mating=2,
                       sol_per_pop=3,
                       num_genes=4,
                       fitness_func=fitness_function,

                       init_range_low=5,
                       init_range_high=15,args=(5,))

What I expected from this piece of code was that I'd print 15 every time the fitness function would be called (just to make sure passing parameters was working correctly).

but instead I get

 python gaex.py
Traceback (most recent call last):
  File "gaex.py", line 14, in <module>
    init_range_high=15,arg=(5,))
TypeError: __init__() got an unexpected keyword argument 'args'

Any suggestions?

Upvotes: 2

Views: 2377

Answers (2)

Chillston
Chillston

Reputation: 301

I found that the provided solution does not work (anymore?) for a parallel execution of the GA, i.e. using the parallel_processing argument. This is because the local function cannot be pickled. Instead one can use global variables:

import pygad

# initialize global attributes
arg1, arg2, ... = None, None, ...

# define a fitness function
def fitness_func(ga_instance, solution, solution_idx):
    global arg1, arg2, ...
    ...
    return fitness

# assign values to the global arguments
arg1, arg2, ... = ...

ga_instance = pygad.GA(
    ...
    fitness_func = fitness_func,
    parallel_processing=['process', 12]
    ...
)

I tried a solution with partial which does seem to work but later shows a weird bug (genes in the population change to values in [-1e-35, 1e-35]).

Upvotes: 0

psarka
psarka

Reputation: 1842

Make your fitness function parametrizable like that:

def fitness_function_factory(num):

    def fitness_function(solution, solution_idx):
        print(num + 10)
        return sum(solution)

    return fitness_function

Then give it to the GA it like that:

ga_instance = pygad.GA(num_generations=1,
                       ...
                       fitness_func=fitness_function_factory(5),
                       ...
                       init_range_high=15)

Upvotes: 4

Related Questions