def simulate(samples, parallel):
# create simulators
simulators = {\
TRProblem: {},
SphereProblemOriginR1: {},
SphereProblemOriginR2: {},
SchwefelsProblem26: {}
}
for problem in problems:
for optimizer in optimizers[problem]:
simulators_op = []
for i in range(0, samples):
simulator = Simulator(optimizer(), problem(), termination)
simulators_op.append(simulator)
simulators[problem][optimizer] = simulators_op
simulate = lambda simulator : simulator.simulate()
# run simulators
for problem in problems:
for optimizer, simulators_ in simulators[problem].iteritems():
if(parallel):
resulting_simulators = pmap(simulate, simulators_)
else:
resulting_simulators = map(simulate, simulators_)
from evopy.operators.termination.accuracy import Accuracy
from evopy.operators.termination.generations import Generations
from evopy.operators.termination.convergence import Convergence
from evopy.external.playdoh import map as pmap
from os.path import exists
from os import mkdir
from setup import *
# create simulators
for problem in problems:
for optimizer in optimizers[problem]:
simulators_op = []
for i in range(0, samples):
simulator = Simulator(optimizer(), problem(), termination)
simulators_op.append(simulator)
simulators[problem][optimizer] = simulators_op
simulate = lambda simulator : simulator.simulate()
# run simulators
for problem in problems:
for optimizer, simulators_ in simulators[problem].iteritems():
resulting_simulators = pmap(simulate, simulators_)
for simulator in resulting_simulators:
cfc = simulator.logger.all()['count_cfc']
cfcs[problem][optimizer].append(cfc)
if not exists("output/"):
mkdir("output/")
sklearn_cv = SVCCVSkGridLinear(\
C_range = [2 ** i for i in range(-5, 5, 2)],
cv_method = KFold(20, 5))
meta_model = DSESSVCLinearMetaModel(\
window_size = 10,
scaling = ScalingStandardscore(),
crossvalidation = sklearn_cv,
repair_mode = 'mirror')
method = ORIDSESAlignedSVC(\
mu = 15,
lambd = 100,
theta = 0.3,
pi = 70,
initial_sigma = matrix([[4.5, 4.5]]),
delta = 4.5,
tau0 = 0.5,
tau1 = 0.6,
initial_pos = matrix([[10.0, 10.0]]),
beta = 0.9,
meta_model = meta_model)
return method
if __name__ == "__main__":
problem = TRProblem()
optfit = problem.optimum_fitness()
sim = Simulator(get_method(), problem, Accuracy(optfit, 10**(-6)))
results = sim.simulate()
