def ga_fitted():
np.random.seed(10)
clf = RandomForestClassifier(max_depth=3, random_state=2222)
ga = GeneticAlgorithm(clf, 5, duration=0.5)
iris = load_iris()
ga.fit(iris.data, iris.target)
return ga
def __init__(self, neural_network, population_size=100, number_of_generations=10, mutation_probability=0.01, number_of_elites=0): """ Initialization function of the class ... Parameters ---------- Specified in the class docstring Returns ------- None Raises ------ None """ # Set the Neural Network self.neural_network = neural_network # Set the default range self.output_range = [-5, 5] # Chromosome length is derived from Neural Network chromosome_length = self.neural_network.number_of_parameters GeneticAlgorithm.__init__(self, population_size, number_of_generations, mutation_probability, chromosome_length, number_of_elites) # Generate some class variables self._chromosome_setting()
def __init__(self, args):
super().__init__(args)
self._name = "q_MX_minfit_p15"
self.max_uid = 0
self._num_simulations = 5
population_size = 30
self.ga = GeneticAlgorithm(population_size,
r_mutation=0.05,
apex_stddev=0.25)
# self.ga.generation = 60
self.result_map = {}
self.points_map = {}
self.alive_map = {}
# population_size = args.population_size
# args.r_mutation = 0.4
self._generation_info = []
for i in range(self.ga.population_size):
self._add_snake()
self._load_genomes()
def simulate(params, target_sentence="Hello World"):
np.random.seed(123)
random.seed(123)
MAX_GEN = params['max_gen'] # termination
MAX_SUCCESS = params['max_success'] # termination
POPULATION_SIZE = params['population_size']
NUM_ATTRIBUTES = len(target_sentence)
MUTATION_PROB = params['mutation_prob']
CROSSOVER_PROB = params['crossover_prob']
GA = GeneticAlgorithm(fitness_function,
num_attributes=NUM_ATTRIBUTES,
population_size=POPULATION_SIZE,
mutation_prob=MUTATION_PROB,
crossover_prob=CROSSOVER_PROB)
GA.initialize_population()
scores = []
generation_counter = 0
success_counter = 0
while generation_counter < MAX_GEN and success_counter < MAX_SUCCESS:
GA.compute_fitness_score()
scores.append(np.mean(GA.fitness_scores))
print('Generation', generation_counter, ", avg score:",
scores[generation_counter], ", best:", GA.get_best())
if GA.get_best() == target_sentence:
success_counter += 1
GA.run()
generation_counter += 1
return scores, generation_counter
def __init__(self,
clf,
fold,
duration=None,
max_iter=None,
base_included=True):
"""Init method.
Args:
clf : classifier object implementing 'fit'
Classfier used for scoring new features.
fold : int, cross-validation generator or an iterable
Determines the cross-validation splitting strategy,
see also http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.cross_val_score.html.
duration : int
Determines how many minutes a genetic algorithm runs.
max_iter : int
Determines how many iterations a genetic algorithm runs.
base_included : bool
Determines whether or not the base dataset is included during the evaluation of newly created features.
ga : GeneticAlgorithm
Object used for creating new sets of features.
"""
self.clf = clf
self.fold = fold
self.duration = duration
self.max_iter = max_iter
self.base_included = base_included
self.ga = GeneticAlgorithm(clf, fold, duration, max_iter,
base_included)
def main():
parser = argparse.ArgumentParser(
description='Poorly solve the general assignment problem.',
fromfile_prefix_chars='@')
parser.add_argument('filename', metavar='filename', type=unicode,
help='a data file to parse and run on')
parser.add_argument('--population', metavar='population', type=int,
default=100, help='how large to keep population')
parser.add_argument('--evaluations', metavar='evaluations', type=int,
default=100000,
help='how many evaluations to run between restarts')
parser.add_argument('--restarts', metavar='restarts', type=int,
default=10,
help='how often to restart from a random point')
args = parser.parse_args()
POPULATION = args.population
NUMBER_OF_EVALUATIONS = args.evaluations
RESTARTS = args.restarts
total_evaluations = 0
top_genotypes = []
for n in xrange(RESTARTS):
print "restart", n, "out of {0}:".format(RESTARTS)
print
with open(args.filename) as f:
ga = GeneticAlgorithm.from_file(f, fittest_found_callback)
ga.generate_random_solutions(POPULATION)
evolutions = NUMBER_OF_EVALUATIONS / POPULATION
for _ in xrange(evolutions):
ga.evolve_population()
indexes = agent_indexes(ga.agents)
top_genotypes.extend(assignment_indexes(solution)
for solution in ga.solution_pool)
total_evaluations += ga.evaluations
with open(args.filename) as f:
ga = GeneticAlgorithm.from_file(f, fittest_found_callback)
ga.solution_pool = [Solution(ga.agents,
assignment_agents(genotype, ga.agents))
for genotype in top_genotypes]
while len(ga.solution_pool) > 10:
ga.double_population()
ga.halve_population()
ga.halve_population()
print "population:", len(ga.solution_pool)
print "all done:"
print "total_evaluations: ", (ga.evaluations + total_evaluations)
best_solution = sorted(ga.solution_pool, key=lambda s: s.total_cost)[0]
fittest_found_callback(best_solution, ga)
def run():
'''
Main program.
'''
args = get_args()
rospy.init_node('pf', anonymous=True)
# Load input
sampling_points = np.genfromtxt(args['rewards'],
delimiter=' ',
dtype=float)
print 'Starting Genetic Algorithm...'
# Run the GA heuristic to get robot path.
ga_op = GeneticAlgorithm(sampling_points, args['max_cost'], args['gen'],
args['pop'])
path, cost, rewards = ga_op.run(
(args['start_x'], args['start_y']),
(args['end_x'], args['end_y']),
)
print 'Done.'
dim = sampling_points.shape
filename = 'path_{}x{}_maxc{}.txt'.format(dim[0], dim[1], args['max_cost'])
with open(filename, 'w') as output_file:
output_file.write('{}\n'.format(args['max_cost']))
output_file.write('{}\n'.format(round(cost, 2)))
output_file.write('{}\n'.format(round(rewards, 2)))
for coord in path:
output_file.write('{}\n'.format(coord))
print_path(dim[0], dim[1], sampling_points, path, cost, rewards,
args['max_cost'])
if args['navigate'] == 0:
return
robot = Robot(RATE, QUEUE_SIZE)
while not robot.is_ready():
pass
print 'Starting navigation through path:', path
goal = None
# Control robot through the path.
while not rospy.is_shutdown() and len(path) > 0:
if goal is None or not robot.bug2(goal[0], goal[1]):
goal = path.pop(0)
print 'Done.'
return
def test_ga_init():
clf = RandomForestClassifier(max_depth=3, random_state=2222)
with pytest.raises(ValueError):
ga = GeneticAlgorithm(clf, 5)
with pytest.raises(ValueError):
ga = GeneticAlgorithm(clf, 5, duration='s')
with pytest.raises(ValueError):
ga = GeneticAlgorithm(clf, 5, max_iter=0.5)
with pytest.raises(ValueError):
ga = GeneticAlgorithm(clf, 5, 1, 1)
with pytest.raises(ValueError):
ga = GeneticAlgorithm(clf, 5, base_included='s')
def train_ga(self, generations=90):
flattened = self.weights_flattened()
initial_solutions = np.random.uniform(low=-1.0,
high=1.0,
size=(20, len(flattened)))
initial_solutions[0] = np.array(flattened, dtype=np.float64)
ga = GeneticAlgorithm(solutions=initial_solutions,
num_parents_for_mating=4,
generations=generations,
fitness_func=self.fitness_func,
offspring_sz=4)
ga.start()
self.weights = self.weights_unflattened(
ga.solutions[0]) #best solution
def main():
# agent params
params = {
"size_pop": 100,
"crossover_rate": 0.9,
"mutation_rate": 0.03,
"generations": 100,
"agent_dimension": 10,
"agent": Rastrigin
}
genetic_algorithm = GeneticAlgorithm(**params)
solution = genetic_algorithm.start()
print(solution)
genetic_algorithm.plotGraphic()
def iniciar():
calcular = CalcFitness()
calcular.precision = 10
# Los coeficientes de z
calcular.get_limites()
calcular.get_mj()
poblacion = Poblacion(10, True, calcular)
algorithm = GeneticAlgorithm(True)
i = 0
print(calcular.mj)
print(calcular.aj)
print(calcular.bj)
print(calcular.longitud)
for i in range(5000):
print('Poblacion: ' + str(i+1))
print(poblacion)
poblacion = algorithm.get_next_generation(poblacion)
print('Poblacion: ' + str(i + 1))
print(poblacion)
def testWithFakeFitness(table_len=1000,
pop_size=100,
chr_size=10,
generations_to_run=200,
random_seed=None):
np.random.seed(random_seed)
fake_fitness_table = np.random.randn(table_len)
def fake_fitness_function(chromossome):
return np.mean(fake_fitness_table[chromossome])
gene_set = list(range(table_len))
selection = Selection(0.5)
crossover = Crossover(kind=1)
mutation = Mutation(0.5)
optimizer = GeneticAlgorithm(gene_set, selection, crossover, mutation)
optimizer.initializePopulation(chr_size, pop_size)
random_optimizer = RandomSearch(gene_set)
random_optimizer.initializePopulation(chr_size, pop_size)
plt.grid()
for i in range(generations_to_run):
optimizer.step(fake_fitness_function)
random_optimizer.step(fake_fitness_function)
plt.scatter([i] * pop_size,
[fake_fitness_function(x) for x in optimizer.population],
c='b')
plt.scatter(i, random_optimizer.best_fitness, c='g')
plt.show()
def ga():
np.random.seed(10)
iris = load_iris()
clf = RandomForestClassifier(max_depth=3, random_state=2222)
ga = GeneticAlgorithm(clf, cv=5, duration=0.5)
ga.X = np.asarray(iris.data)
ga.y = np.asarray(iris.target)
ga.y = ga.y.reshape(ga.y.shape[0], )
ga.n_features = np.random.random_integers(10)
return ga
class GeneticAlgorithmTestCase(unittest.TestCase):
def setUp(self):
self.ga = GeneticAlgorithm(threads=1, env_name="BipedalWalker-v3", max_episode_len=100, seed=42)
def test_smoke_mutate(self):
sigma = 0.002
ind = self.ga.init_population(1)[0]
mutated = self.ga.mutate(ind, sigma)
def test_smoke_evalute_fitness(self):
ind = self.ga.init_population(1)[0]
fitness = self.ga.evaluate_fitness(ind)
@async_test
async def test_smoke_async_evaluate_fitness(self):
ind = self.ga.init_population(1)[0]
task = asyncio.gather(
asyncio.ensure_future(self.ga.evaluate_fitness(ind)),
asyncio.ensure_future(self.ga.evaluate_fitness(ind)))
await task
def testWithRealFitness(fitness_function,
table_len,
pop_size=100,
chr_size=10,
generations_to_run=200,
random_seed=None):
gene_set = list(range(table_len))
selection = Selection(0.5)
crossover = Crossover(kind=1)
mutation = Mutation(0.5)
optimizer = GeneticAlgorithm(gene_set, selection, crossover, mutation)
optimizer.initializePopulation(chr_size, pop_size)
random_optimizer = RandomSearch(gene_set)
random_optimizer.initializePopulation(chr_size, pop_size)
plt.grid()
for i in range(generations_to_run):
optimizer.step(fitness_function)
random_optimizer.step(fitness_function)
plt.scatter([i] * pop_size,
[fitness_function(x) for x in optimizer.population],
c='b')
plt.scatter(i, random_optimizer.best_fitness, c='g')
plt.show()
return optimizer
from ga import GeneticAlgorithm ga = GeneticAlgorithm(elitism=False) ga.run(1500)
class FeatureConstructor:
"""Create new features using genetic algorithm."""
def __init__(self,
clf,
fold,
duration=None,
max_iter=None,
base_included=True):
"""Init method.
Args:
clf : classifier object implementing 'fit'
Classfier used for scoring new features.
fold : int, cross-validation generator or an iterable
Determines the cross-validation splitting strategy,
see also http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.cross_val_score.html.
duration : int
Determines how many minutes a genetic algorithm runs.
max_iter : int
Determines how many iterations a genetic algorithm runs.
base_included : bool
Determines whether or not the base dataset is included during the evaluation of newly created features.
ga : GeneticAlgorithm
Object used for creating new sets of features.
"""
self.clf = clf
self.fold = fold
self.duration = duration
self.max_iter = max_iter
self.base_included = base_included
self.ga = GeneticAlgorithm(clf, fold, duration, max_iter,
base_included)
def fit(self, X, y):
"""Fit estimator.
Args:
X : array-like
The data to fit.
y : array-like
The target variable.
"""
self.ga.fit(X, y)
def get_params(self, ind='best'):
"""Print best or most frequent set of new features.
Args:
ind : string, 'best' or 'most_freq'
Determines which set of features save to a file.
"""
self.ga.get_params(ind)
def save(self, filename, ind='best'):
"""Save the best or most frequent set of features to a file.
Args:
filename : string
ind : string, 'best' or 'most_freq'
Determines which set of features save to a file.
"""
if ind == 'best' or ind == 'most_freq':
self.ga.save(filename, ind)
else:
raise ValueError("ind must be 'best' or 'most_freq'.")
def load(self, filename):
"""Load a set of features from a file.
Args:
filename : string
Returns:
Tuple with a set of features.
"""
return self.ga.load(filename)
def transform(self, X, individual):
"""Transform dataset into new one using created features.
Args:
X : array-like
The data to transform.
individual : tuple
Tuple with a set of features.
Returns:
New dataset, array-like.
"""
return self.ga.transform(X, individual)
def plot(self):
"""Plot data from the genetic algorithm."""
self.ga.plot()
def main():
GA = GeneticAlgorithm(Settings.POPULATION_SIZE)
GA.evolve(Settings.MAX_GENERATIONS)
break
try:
chromosome = re.findall(r'[0-9,]+', line)[0]
except IndexError:
logger2.error("Check your {} chromosome!!".format(string))
sys.exit(0)
chromosome_bits = chromosome.split(',')
# convert str to int
buy_chromosome_bits = [int(x) for x in chromosome_bits]
#(4.) create solution
s = Solution()
s.chromosome_bits = buy_chromosome_bits
#(5.) compute fitness
ga = GeneticAlgorithm(parameter_dict, testing_data_dict)
# -(a) translate
s.translate_chromosome_bits(ga.feature_pos_dict)
# -(c) get the classfiled result in each day
classification_result = s.get_classification_result(ga)
if not classification_result:
print("No stocks returned for {} chromosome".format(string))
continue
# -(d) compute fitness
american_stock_fitness = AmericanStockFitness(parameter_dict)
american_stock_fitness(testing_data_dict, s)
testing_output_path = "test_data_result/{}_chromosome_testing.txt".format(
string)
print("testing_output_path: ", testing_output_path)
Solution.compute_profit()
import time
from tqdm import tqdm
import matplotlib.pyplot as plt
from ga import GeneticAlgorithm
if __name__ == "__main__":
phrase = "fire ball" # The phrase we want the computer to match
search_space = list(
"abcdefghijklmnopqrstuvwxyz "
) # Our search space (to tell the computer its possible options)
g_a = GeneticAlgorithm(search_space)
# Perform experiment.
run_iters = []
run_time = []
exp_start = time.time()
for i in tqdm(range(10)):
start_time = time.time()
curr_iter = g_a.run(phrase)
time_taken = time.time() - start_time
# Append iterations taken.
run_iters.append(curr_iter)
# Append time taken.
run_time.append(time_taken)
print(
f"Total time taken to run experiment: {time.time() - exp_start:.2f} seconds"
)
# seed radius
IS = parameter_dict['DSGA']['IS']
# radius delta
SD = parameter_dict['DSGA']['SD']
seed_radius = SeedRadius(parameter_dict)
# (2.) put data into dict
formatter1 = Formatter(parameter_dict)
input_data_dict = formatter1.format_and_create_dict(
formatter1.path, formatter1.feature_choice_list)
logger1.info("create input_data_dict successful")
# get the range of the feature value
formatter1.compute_chosen_feature_value_range()
# (3.) create initial parents
ga = GeneticAlgorithm(parameter_dict, input_data_dict)
ga.seed_radius = seed_radius
ga.create_initial_parents()
off_spring_generation = OffspringGeneration(parameter_dict)
big_loop = 100
while not ga.END:
RLC = parameter_dict['DSGA']['RLC']
for i in range(50):
# (4.) offspring generation , return target, compute fitness
current_solution_pool = off_spring_generation(Solution.all())
ga.process_new_solutions(current_solution_pool)
# (5.) compute shared fitness
Solution.compute_shared_fitness(ga)
# (6.) find seed solution
Solution.find_seed_solution(ga)
pipes.pipe_up) == 1:
return True
elif bird.rect.y < 0 or bird.rect.y > cs.HEIGHT:
return True
return False
def get_closest_pipe(piepes, x):
for pipe in piepes:
if pipe.pipe_down.x - x + cs.P_WIDTH > 0:
return pipe
firstStart = True
gen = GeneticAlgorithm()
def play(daemon, firstStart=False, generation=1):
counter = 0
if firstStart:
population = gen.population
else:
population = gen.crossover()
fitness = 0
deadcounter = 0
pipes = [Pipe(randint(120, 400))]
from env_ones import Ones
from ga import GeneticAlgorithm
e = Ones()
par = {"crossover.type" : "single",
"elitism" : False,
"n.generations" : 100,
"n.individuals" : 100,
"p.crossover" : 0.8,
"p.mutation" : 0.01,
"selection.type" : "tournament.selection",
"tournament.size" : 2,
"type" : "simple"}
g = GeneticAlgorithm(e, par)
print (g.population)
g.evolve()
print (g.population)
prob_weight_reset = 0.001
max_gen = 500
weight_range = 50.0
mut_range = 5.0
target_values = [10.0, 20.0]
ga = GeneticAlgorithm(len_indv=len_indv,
pop_size=pop_size,
num_par=num_par,
prob_mut=prob_mut,
prob_xover=prob_xover,
prob_survival=prob_survival,
prob_weight_reset=prob_weight_reset,
max_gen=max_gen,
weight_range=weight_range,
mut_range=mut_range,
target_values=target_values,
node_types=node_types,
n_nodes_input=n_nodes_input,
n_nodes_hidden=n_nodes_hidden,
n_nodes_output=n_nodes_output,
add_bias=add_bias,
out_ma_len=out_ma_len,
tau=tau,
weight_epsilon=weight_epsilon,
max_ticks=max_ticks)
ga.evolve()
# Housekeeping to save all pertinent experiment results:
script_path = os.getcwd()
exp_path = script_path + '\\' + exp_id
}
"""
"""
par = {"crossover.type" : "single",
"elitism" : False,
"n.generations" : 100,
"n.individuals" : 100,
"p.crossover" : 0.8,
"p.mutation" : 0.01,
"selection.type" : "tournament.selection",
"tournament.size" : 2,
"type" : "simple",
"sharing.domain" : "phenotype",
"alpha.share" : 1,
"theta.share" : 0.1,
"fitness.sharing" : True
}
"""
prom = tsp.energiaPromedio()
sa = SimulatedAnnealing(tsp)
best = list(sa.search(0.999, prom, prom / 1000, 100)[0])
tsp.showSA(best)
tsp = TSP(20, True)
ga = GeneticAlgorithm(tsp, par)
tsp.show(ga.population)
print(ga.population)
ga.evolve()
tsp.show(ga.population)
features_cols = train_x[0].shape[0]
features_rows = train_x[0].shape[1]
parameter_size = train_y[0].shape[0]
features = tf.placeholder(tf.float32, [None, features_cols, features_rows])
patches = tf.placeholder(tf.float32, [None, parameter_size])
prob_keep_input = tf.placeholder(tf.float32)
prob_keep_hidden = tf.placeholder(tf.float32)
batch_size = tf.placeholder(tf.int32)
warnings.simplefilter("ignore")
lstm = LSTM(features=features, labels=patches, batch_size=batch_size)
ga = GeneticAlgorithm(extractor=extractor, population_size=200,
percent_elitism_elites=5, percent_elitist_parents=5,
dna_length=(parameter_size), target_features=test_x,
feature_size=(features_cols * features_rows),
mutation_rate=0.01, mutation_size=0.1)
hill_climber = HillClimber(extractor=extractor, target_features=test_x,
feature_size=(features_cols * features_rows),
parameter_size=parameter_size,
averaging_amount=4)
mlp = MLP(features=features, labels=patches, parameters=[50, 40, 30],
prob_keep_input=prob_keep_input,
prob_keep_hidden=prob_keep_hidden)
if parameter_size == 155:
hier_mlp = RecursiveMLP(features=features, labels=patches,
parameters=[50, 40, 30],
target_values[0, 0] = 5.0
target_values[0, 1] = 15.0
#target_values = np.array([5.0, 10.0])
ga = GeneticAlgorithm(len_indv=len_indv,
pop_size=pop_size,
num_par=num_par,
prob_mut=prob_mut,
prob_xover=prob_xover,
prob_survival=prob_survival,
prob_weight_reset=prob_weight_reset,
max_gen=max_gen,
weight_range=weight_range,
mut_range=mut_range,
target_values=target_values,
node_types=node_types,
n_nodes_input=n_nodes_input,
n_nodes_cortex=n_nodes_cortex,
n_nodes_hidden=n_nodes_hidden,
n_nodes_output=n_nodes_output,
intralayer_connections_flag=intralayer_connections_flag,
add_bias=add_bias,
out_ma_len=out_ma_len,
tau=tau,
max_ticks=max_ticks,
switch_tick=switch_tick,
bg_nodes=bg_nodes,
bg_sync_freq=bg_sync_freq)
ga.evolve()
archive_results(exp_id, ga)
table = randtable(n)
tc = trueclauses(table, formula)
if tc > local:
local = tc
if local > randsearch:
randsearch = local
print(randsearch)
return randsearch
formula, n = readsat("uf100-01.cnf")
#print(formula, len(formula))
popsize = 60
epochs = 10000
print("# Global Optimum:", len(formula))
#print("randsearch", randomsearch(popsize, epochs, formula, n))
ga = GeneticAlgorithm(n,
crossoverrate=0.99,
mutationrate=(0.05, 0.01),
popsize=popsize,
elitism=5,
epochs=epochs)
ga.setfitness(fitness, formula)
ga.setgenometophenome(gtop)
ga.setphenometogenome(ptog)
solucao = ga.evolve()
print(solucao)
#Add a column of random numbers to the data
rand_col_verif = 100 * np.random.rand(len(x_verif))
x_verif = np.insert(x_verif, 2, rand_col_verif, axis=1)
"""
Segment 3: GA, I guess.
"""
#Run GA to find best weights
N_init_pop = 50
N_crossover = 100
N_selection = 50
improv_thresh = 1e-3
print("Step 1.")
weight_ga = GeneticAlgorithm(3, N_init_pop, mu=0.1)
weight_pop = weight_ga.get_population()
metric_array = np.empty(N_init_pop)
for i in range(len(weight_pop)):
#Scale input data
scaled_x_train = np.multiply(x_train, weight_pop[i])
#Scale verificaion data
scaled_x_verif = np.multiply(x_verif, weight_pop[i])
#Method 1
reg = KNNRegressor(scaled_x_train, y_train, 5)
neighbors = reg.find_all_neighbors(scaled_x_verif)
nbh_std = reg.find_neighborhood_std(neighbors)
metric_array[i] = nbh_std
#Update fitness in GA object
# ====================format ga_unittest_dict end=======================
#-----------------------get targets returned and fitness
#(1.) read para
reader1 = ReadParameters()
parameter_dict = reader1.read_parameters(para_file_path)
# (2.) put data into dict
formatter1 = Formatter(parameter_dict, path=data_file_path)
input_data_dict = formatter1.format_and_create_dict(
formatter1.path, formatter1.feature_choice_list)
formatter1.compute_chosen_feature_value_range()
# (3.) create initial parents
ga = GeneticAlgorithm(parameter_dict, input_data_dict)
#(4.) create solution
s = Solution()
s.chromosome_bits = ga_unittest_dict['chromosome_bits']
# (a) translate chromosome bits list to decimal value
s.translate_chromosome_bits(ga.feature_pos_dict)
# (b) get the classfiled result in each day
s.get_classification_result(ga)
# (c) compute the fitness for solution
american_stock_fitness = AmericanStockFitness(parameter_dict)
american_stock_fitness(ga.input_data_dict, s)
# TESING CODE
golden_return = ga_unittest_dict['golden_result']
golden_fitness = ga_unittest_dict['golden_fitness']
def ga_run(x_train, y_train, x_test, y_test, x_verif, y_verif, k):
# Run GA to find best weights.
N_init_pop = 50
N_crossover = 50
N_selection = 20
improv_thresh = 1e-3
_, nFeats = np.shape(x_train)
weight_ga = GeneticAlgorithm(nFeats, N_init_pop, mu=0.1)
weight_pop = weight_ga.get_population()
metric_array = np.empty(N_init_pop)
# Create the initial population.
for i in range(len(weight_pop)):
# Scale input data
scaled_x_train = np.multiply(x_train, weight_pop[i])
# Scale verificaion data
scaled_x_verif = np.multiply(x_verif, weight_pop[i])
# Regressor.
reg = KNNRegressor(scaled_x_train, y_train, k)
neighbors = reg.find_all_neighbors(scaled_x_verif)
nbh_std = reg.find_neighborhood_std(neighbors)
metric_array[i] = nbh_std
# Update fitness in GA object.
weight_ga.set_fitness(metric_array)
weight_ga.selection(N_selection)
new_best_metric = 2.5
# while (best_metric - new_best_metric) > improv_thresh:
count = 0
while (count < 20):
count += 1
best_metric = new_best_metric
# Crossover.
weight_ga.crossover(N_crossover)
# Get new population.
weight_pop = weight_ga.get_population()
metric_array = np.empty(N_crossover)
# Evaluate and set fitness.
for i in range(len(weight_pop)):
# Scale input data
scaled_x_train = np.multiply(x_train, weight_pop[i])
# Scale verificaion data
scaled_x_verif = np.multiply(x_verif, weight_pop[i])
# Regressor.
reg = KNNRegressor(scaled_x_train, y_train, k)
neighbors = reg.find_all_neighbors(scaled_x_verif)
nbh_std = reg.find_neighborhood_std(neighbors)
metric_array[i] = nbh_std
# Update fitness in GA object
weight_ga.set_fitness(metric_array)
# get_best_sol
best_weights, new_best_metric = weight_ga.best_sol()
#print("Metric of this iteration are: ", new_best_metric)
weight_ga.selection(N_selection)
# print("Best weights = ", best_weights, "\tBest metric = ", new_best_metric)
# Test with scaling after GA
# Concatenate training and verification sets.
x_train = np.concatenate((x_train, x_verif), axis=0)
y_train = np.concatenate([y_train, y_verif])
# Print the results of KNN.
reg = KNNRegressor(np.multiply(x_train, best_weights), y_train, k)
y_pred = reg.predict(np.multiply(x_test, best_weights))
mse_iter = skmse(y_test, y_pred)
print("ga,knn,", k, ",", mse_iter)
# Print the results of KNN.
reg = DwKNNRegressor(np.multiply(x_train, best_weights), y_train, k)
y_pred = reg.predict(np.multiply(x_test, best_weights))
mse_iter = skmse(y_test, y_pred)
print("ga,dknn,", k, ",", mse_iter)