def test_BA_iters_to_fes(self):
task = Task(D=10,
nGEN=1000,
optType=OptimizationType.MINIMIZATION,
benchmark=Sphere())
algo = BatAlgorithm(task=task, NP=10)
algo.run()
evals = algo.task.evals()
self.assertEqual(evals, 10010)
def test_BA_iters_fine(self):
task = Task(D=10,
nGEN=1000,
optType=OptimizationType.MINIMIZATION,
benchmark=Sphere())
algo = BatAlgorithm(task=task, NP=25)
algo.run()
iters = algo.task.iters()
self.assertEqual(iters, 1000)
def test_BA_evals_fine(self):
task = Task(D=10,
nFES=1000,
optType=OptimizationType.MINIMIZATION,
benchmark=Sphere())
algo = BatAlgorithm(task=task, NP=25)
algo.run()
evals = algo.task.evals()
self.assertEqual(evals, 1000)
class BATestCase(TestCase):
def setUp(self):
self.ba_custom = BatAlgorithm(10, 20, 10000, 0.5, 0.5, 0.0, 2.0,
MyBenchmark())
self.ba_griewank = BatAlgorithm(10, 40, 10000, 0.5, 0.5, 0.0, 2.0,
'griewank')
def test_custom_works_fine(self):
self.assertTrue(self.ba_custom.run())
def test_griewank_works_fine(self):
self.assertTrue(self.ba_griewank.run())
def optimize(bench, algo):
average_mfo = 0
average_de = 0
average_abc = 0
average_pso = 0
average_ba = 0
average_fa = 0
average_ga = 0
for i in np.arange(epoch):
mfo = MothFlameOptimizer(D=dim, NP=pop, nGEN=maxIter, benchmark=bench)
de = DifferentialEvolution(D=dim,
NP=pop,
nGEN=maxIter,
benchmark=bench)
abc = ArtificialBeeColonyAlgorithm(D=dim,
NP=pop,
nFES=maxIter,
benchmark=bench)
pso = ParticleSwarmAlgorithm(D=dim,
NP=pop,
nGEN=maxIter,
benchmark=bench)
ba = BatAlgorithm(D=dim, NP=pop, nFES=maxIter, benchmark=bench)
fa = FireflyAlgorithm(D=dim, NP=pop, nFES=maxIter, benchmark=bench)
ga = GeneticAlgorithm(D=dim, NP=pop, nFES=maxIter, benchmark=bench)
gen, best_de = de.run()
gen, best_mfo = mfo.run()
gen, best_abc = abc.run()
gen, best_pso = pso.run()
gen, best_ba = ba.run()
gen, best_fa = fa.run()
gen, best_ga = ga.run()
average_mfo += best_de / epoch
average_de += best_mfo / epoch
average_abc += best_abc / epoch
average_pso += best_pso / epoch
average_ba += best_ba / epoch
average_fa += best_fa / epoch
average_ga += best_ga / epoch
print(algo, ': DE Average of Bests over', epoch, 'run: ', average_de)
print(algo, ': MFO Average of Bests over', epoch, 'run: ', average_mfo)
print(algo, ': ABC Average of Bests over', epoch, 'run: ', average_abc)
print(algo, ': PSO Average of Bests over', epoch, 'run: ', average_pso)
print(algo, ': BA Average of Bests over', epoch, 'run: ', average_ba)
print(algo, ': FA Average of Bests over', epoch, 'run: ', average_fa)
print(algo, ': GA Average of Bests over', epoch, 'run: ', average_ga)
return [
average_de, average_mfo, average_abc, average_pso, average_ba,
average_fa, average_ga
]
class BATestCase(TestCase):
def setUp(self):
self.ba_custom = BatAlgorithm(D=10,
NP=20,
nFES=1000,
A=0.5,
r=0.5,
Qmin=0.0,
Qmax=2.0,
benchmark=MyBenchmark())
self.ba_griewank = BatAlgorithm(NP=10,
D=40,
nFES=1000,
A=0.5,
r=0.5,
Qmin=0.0,
Qmax=2.0,
benchmark='griewank')
def test_custom_works_fine(self):
self.assertTrue(self.ba_custom.run())
def test_griewank_works_fine(self):
self.assertTrue(self.ba_griewank.run())
# encoding=utf8
# This is temporary fix to import module from parent folder
# It will be removed when package is published on PyPI
import sys
sys.path.append('../')
# End of fix
from NiaPy.algorithms.basic import BatAlgorithm
from NiaPy.util import StoppingTask, OptimizationType
from NiaPy.benchmarks import Sphere
# we will run Bat Algorithm for 5 independent runs
for i in range(5):
task = StoppingTask(D=10,
nGEN=1000,
optType=OptimizationType.MINIMIZATION,
benchmark=Sphere())
algo = BatAlgorithm(NP=40, A=0.5, r=0.5, Qmin=0.0, Qmax=2.0)
best = algo.run(task=task)
print('%s -> %s' % (best[0], best[1]))
return evaluate
# example using custom benchmark "MyBenchmark"
logger.info('Running with custom MyBenchmark...')
for i in range(10):
Algorithm = BatAlgorithm(D=10,
NP=40,
nFES=10000,
A=0.5,
r=0.5,
Qmin=0.0,
Qmax=2.0,
benchmark=MyBenchmark())
Best = Algorithm.run()
logger.info(Best)
# example using predifined benchmark function
# available benchmarks are:
# - griewank
# - rastrigin
# - rosenbrock
# - sphere
logger.info('Running with default Griewank benchmark...')
griewank = Griewank()
for i in range(10):
Algorithm = BatAlgorithm(D=10,
NP=40,
class BatAlgorithm(FeatureSelectionAlgorithm):
r"""Implementation of feature selection using BA algorithm.
Date:
2020
Author:
Luka Pečnik
Reference:
The implementation is adapted according to the following article:
D. Fister, I. Fister, T. Jagrič, I. Fister Jr., J. Brest. A novel self-adaptive differential evolution for feature selection using threshold mechanism . In: Proceedings of the 2018 IEEE Symposium on Computational Intelligence (SSCI 2018), pp. 17-24, 2018.
Reference URL:
http://iztok-jr-fister.eu/static/publications/236.pdf
License:
MIT
See Also:
* :class:`niaaml.preprocessing.feature_selection.feature_selection_algorithm.FeatureSelectionAlgorithm`
"""
Name = 'Bat Algorithm'
def __init__(self, **kwargs):
r"""Initialize BA feature selection algorithm.
"""
self._params = dict(A=ParameterDefinition(MinMax(0.5, 1.0),
param_type=float),
r=ParameterDefinition(MinMax(0.0, 0.5),
param_type=float),
Qmin=ParameterDefinition(MinMax(0.0, 1.0),
param_type=float),
Qmax=ParameterDefinition(MinMax(1.0, 2.0),
param_type=float))
self.__ba = BA(NP=10)
def set_parameters(self, **kwargs):
r"""Set the parameters/arguments of the algorithm.
"""
kwargs['NP'] = self.__ba.NP
self.__ba.setParameters(**kwargs)
def __final_output(self, sol):
r"""Calculate final array of features.
Arguments:
sol (numpy.ndarray[float]): Individual of population/ possible solution.
Returns:
numpy.ndarray[bool]: Mask of selected features.
"""
selected = numpy.ones(sol.shape[0] - 1, dtype=bool)
threshold = sol[sol.shape[0] - 1]
for i in range(sol.shape[0] - 1):
if sol[i] < threshold:
selected[i] = False
return selected
def select_features(self, x, y, **kwargs):
r"""Perform the feature selection process.
Arguments:
x (pandas.core.frame.DataFrame): Array of original features.
y (pandas.core.series.Series) Expected classifier results.
Returns:
pandas.core.frame.DataFrame: Mask of selected features.
"""
num_features = x.shape[1]
benchmark = _FeatureSelectionThresholdBenchmark(x, y)
task = StoppingTask(D=num_features + 1, nFES=1000, benchmark=benchmark)
best = self.__ba.run(task)
return self.__final_output(benchmark.get_best_solution())
def to_string(self):
r"""User friendly representation of the object.
Returns:
str: User friendly representation of the object.
"""
return FeatureSelectionAlgorithm.to_string(self).format(
name=self.Name,
args=self._parameters_to_string(self.__ba.getParameters()))
