CuckooSearch's runIteration is incompatible with other algorithms runIteration

[karakatic]

CuckooSearch's runIteration is incompatible with other algorithms' runIteration. It is because of the required argument pa_v. This makes it impossible to use it in the experiments with generic method calls.

NiaPy/NiaPy/algorithms/basic/cs.py

Line 135 in a135cc0

def runIteration(self, task, pop, fpop, xb, fxb, pa_v, **dparams):

[karakatic]

To expand this issue: many NIA optimizers were changed in the latest version to extend the standard required arguments when calling runIteration. Earlier versions forced all the NIA parameter settings at the construction, not in the runIteration call... This change makes NiaPy unusable in other frameworks, where generic runIteration calls are made...

[sisco0]

Have a good day @karakatic . Would you recommend to move out these parameters to constructor function instead of being at the runIteration function?

[kb2623]

This parameter should not be and issue, because in Algorithm interface, method runYield unpacks the parameter as seen from:

NiaPy/NiaPy/algorithms/algorithm.py

Line 296 in a135cc0

pop, fpop, xb, fxb, dparams = self.runIteration(task, pop, fpop, xb, fxb, **dparams)

Method runIteration has an additional argument, because I mead the code shorter. If there was no additional argument, algorithm developer should manually obtain this parameter from argument dparams.

Many advanced IDEs report this as an issue, as for some other incompatible type problems, but Python code still runs.

Parameter pa_v is set when run is started, seen from:

NiaPy/NiaPy/algorithms/basic/cs.py

Line 132 in a135cc0

d.update({'pa_v': self.NP * self.pa})

If this parameter is set when constructing an instance of CuckooSearch class, then the code for updating parameters (method setParametes) should be updated. But this parameter is not a default parameter of Cuckoo Search algorithm, based on article Yang, Xin-She, and Suash Deb. "Cuckoo search via Lévy flights." Nature & Biologically Inspired Computing, 2009. NaBIC 2009. World Congress on. IEEE, 2009, so I did not set this parameter at creation of Cuckoo Search algorithm instance.

[zStupan]

IDEs report this as a problem because it seemingly goes against the Liskov substitution principle of object oriented programming. But it actually doesn't because of keyword arguments. Also, runIteration is essentially a private/protected method, I see no use case where you would need to call runIteration directly in your code.

In this particular case though, I don't think pa_v should be an additional parameter, because it stays the same through all iterations. It should just be set in the constructor/setParameters method. If it's optional there should be a flag in setParameters, something like use_pa_v=True and if it's true, self.pa should be set to NP * pa.

@karakatic Can you provide an example of this causing problems?

If it's an issue we would just need to change every algorithms runIteration to pop the additional parameters from dparams like I did in my BFOA implementation:

NiaPy/NiaPy/algorithms/basic/bfoa.py

Lines 193 to 214 in 0fa51cb

	def runIteration(self, task, pop, fpop, xb, fxb, **dparams):
	r"""Core function of Bacterial Foraging Optimization algorithm.
	Args:
	task (Task): Optimization task.
	pop (numpy.ndarray): Current population.
	fpop (numpy.ndarray): Current populations fitness/function values.
	xb (numpy.ndarray): Global best individual.
	fxb (float): Global best individuals function/fitness value.
	**dparams (Dict[str, Any]): Additional arguments.
	Returns:
	Tuple[numpy.ndarray, numpy.ndarray, numpy.ndarray, float, Dict[str, Any]]:
	1. New population.
	2. New populations function/fitness values.
	3. New global best solution,
	4. New global best solutions fitness/objective value.
	5. Additional arguments.
	"""
	cost = dparams.pop('cost')
	health = dparams.pop('health')

We could also just do this to remove the warnings.

[firefly-cpp]

Fixed in #309

@zStupan please post your solution (current status) and final comments before we close this issue.

[zStupan]

As I described in my previous comment, for the algorithms that had extra arguments in the run_iteration method, I removed them and popped them from the keyword arguments inside run_iteration.

In the case of CS, the pa_v parameter was actually static, it represents the number of nests abandoned each iteration i. e. pa * population_size, where pa is the probability of a nest being abandoned. So I just added an attribute self.num_abandoned in init which stores that value so no additional parameters are actually needed.