Additional parameters to complete the objective function

[dangngohai]

Hello,
Is it possible to add more inputs that are fixed parameters needed to completely specify the objective function (different to parameter need to estimate)?
Thank you very much for sharing your library. It helps me so much.

[dangngohai]

In my opinion, you should define your library under classes. I just have look at L-shade method only. I think it would be easier to pass global variables to different methods and give more flexibility by creating possibility of adjusting inputs rather than using only your default parameters.
For adding extra fixed parameters, I think you should add more extra arbitrary positional inputs (e.g *arg) to pyade.commons.apply_fitness and to optimisation functions in methods (for L-SHADE it is apply function).

[dangngohai]

Another question about DE: Why do you compare the population of current generation to 4 (for example:
in current_to_pbest_mutation: if len(population) < 4: return population
or in L-SHADE: # Adapt population size new_population_size = round((4 - init_size) / max_evals * num_evals + init_size)
Where does '4' stand for? I understand it should be the smallest possible size of current generation? but why it is 4?

[xKuZz]

Currently I don't have enought time to change the entire library to use classes.

Now you can use a new parameter called 'opts' that will be passed to the fitness function in which you can include whatever you. See the simple example at the end of README.md

You can change all of the parameters that the function apply gets by modifying the params dict. There may be some parameters fixed in the code by default to the values used by the author of each algorithm. If that's the case and you need access to some of them, open a new issue and I will work on that when possible.

4 is not the mimimum size of the current generation but the minimum size of a generation during the algorithm (reached at the end). That's because the current-to-pbest mutation, requires at least 4 individuals to work. See Eq. 10 on L-SHADE paper

Thanks for your feedback.

[Kaeryv]

For those who would need to do such a thing: don't forget python's functools.

from functools import partial
def objective_function(x, param1, param2, param3):
    pass

objective_function_cmplt = partial(objective_function, param1=..., param2=..., param3=...)