cancer_data['symmetry_mean'], cancer_data['fractal_dimension_mean'],
cancer_data['radius_se'], cancer_data['texture_se'],
cancer_data['perimeter_se'], cancer_data['area_se'],
cancer_data['smoothness_se'], cancer_data['compactness_se'],
cancer_data['concavity_se'], cancer_data['concave_points_se'],
cancer_data['symmetry_se'], cancer_data['fractal_dimension_se'],
cancer_data['radius_worst'], cancer_data['texture_worst'],
cancer_data['perimeter_worst'], cancer_data['area_worst'],
cancer_data['smoothness_worst'], cancer_data['compactness_worst'],
cancer_data['concavity_worst'], cancer_data['concave_points_worst'],
cancer_data['symmetry_worst'], cancer_data['fractal_dimension_worst']))
x = x / np.max(x, axis=0)
y = np.array([float(item == b'B')
for item in cancer_data['diagnosis']]).reshape(len(x), 1)
with utils.OutSplit('breast-cancer'):
population_count = 15
population_size = 15
node_cap = 100
generations = 50
target_accuracy = 0.97
r = 0
np.random.seed(r)
perm = np.random.permutation(len(x))
#np.random.shuffle(x)
#np.random.shuffle(y)
x = x[perm]
y = y[perm]
import numpy as np
import evolutionary_module as evm
import utils
np.random.seed(62)
with utils.OutSplit('multiple_sin') as out_split:
x = (np.random.random(50000)*2*np.math.pi).reshape(50000, 1)
y = (np.sin(x)+1).reshape(50000, 1)
val_x = (np.random.random(1000)*2*np.math.pi).reshape(1000, 1)
val_y = (np.sin(val_x)+1).reshape(1000, 1)
# Grab input from the user and print it to stdout
population_count = int(input('population_count: '))
print('{}'.format(population_count))
population_size = int(input('poplation_size: '))
print('{}'.format(population_size))
node_cap = int(input('node_cap: '))
print('{}'.format(node_cap))
generations = int(input('generations: '))
print('{}'.format(generations))
target_accuracy = float(input('target_accuracy: '))
print('{}%'.format(target_accuracy*100))
r = int(input("random_seed: "))
print(str(r))
for i in range(50):
best = evm.evolve_node_count(x, y, val_x, val_y, utils.jit_near_compare, population_count, population_size, node_cap, generations, target_accuracy, i*r+r)
with utils.OutSplit('nultiple_cos'.format(d= datetime.datetime.now())):
import numpy as np
import evolutionary_module as evm
import utils
np.random.seed(57)
with utils.OutSplit('multiple_xor') as out_split:
x = np.array([[1, 0] if rand < 0.25 else
[0, 1] if rand < 0.5 else [1, 1] if rand < 0.75 else [0, 0]
for rand in np.random.random(60000)],
dtype=int)
y = np.array([[
1,
] if np.sum(itm) == 1 else [
0,
] for itm in x],
dtype=int)
# Grab input from the user and print it to stdout
population_count = int(input('population_count: '))
print('{}'.format(population_count))
population_size = int(input('poplation_size: '))
print('{}'.format(population_size))
node_cap = int(input('node_cap: '))
print('{}'.format(node_cap))
generations = int(input('generations: '))
print('{}'.format(generations))
target_accuracy = float(input('target_accuracy: '))
print('{}%'.format(target_accuracy * 100))
r = int(input("random_seed: "))
print(str(r))
if __name__ == "__main__":
np.random.seed(0)
x = np.array([[1, 0] if rand < 0.25 else
[0, 1] if rand < 0.5 else [1, 1] if rand < 0.75 else [0, 0]
for rand in np.random.random(60000)],
dtype=int)
y = np.array([[
1,
] if np.sum(itm) == 1 else [
0,
] for itm in x],
dtype=int)
with utils.OutSplit('xor_evolution_test'):
# network, fitness, accuracy = evolution(x, y, x, y, utils.jit_round_compare, 30, 4, 50)
# print(fitness, accuracy, network.validate(x, y, utils.jit_round_compare))
best_xor = evolve_node_count(x,
y,
x,
y,
utils.jit_round_compare,
population_count=30,
population_size=15,
node_cap=50,
generations=50,
target_accuracy=1,
r=2)
best_xor.predict(np.array([[0, 1], [1, 1], [0, 0], [1, 0]]))
utils.display_network(best_xor, 'xor')
import numpy as np
import utils
import datetime
import evolutionary_module
np.random.seed(99)
with utils.OutSplit('sin-50k'.format(d=datetime.datetime.now())):
population_count = int(input('population_count: '))
print(population_count)
population_size = int(input('poplation_size: '))
print(population_size)
node_cap = int(input('node_cap: '))
print(node_cap)
generations = int(input('generations: '))
print(generations)
target_accuracy = float(input('target_accuracy: '))
print('{}%'.format(target_accuracy*100))
r = float(input('random_number: '))
print(r)
x = (np.random.random(50000)*2*np.math.pi).reshape(50000, 1)
y = (np.sin(x)+1).reshape(50000, 1)
val_x = (np.random.random(1000)*2*np.math.pi).reshape(1000, 1)
val_y = (np.sin(val_x)+1).reshape(1000, 1)
network = evolutionary_module.evolve_node_count(x, y, val_x, val_y, utils.jit_near_compare, population_count, population_size, node_cap, generations, target_accuracy, r)
import numpy as np
import evolutionary_module as evm
import utils
np.random.seed(57)
with utils.OutSplit('multiple_cancer') as out_split:
cancer_data = np.genfromtxt('datasets\\breast-cancer-wisconsin-data.csv', delimiter=',', dtype=[('id', '<i4'), ('diagnosis', 'S1'), ('radius_mean', '<f8'), ('texture_mean', '<f8'), ('perimeter_mean', '<f8'), ('area_mean', '<f8'), ('smoothness_mean', '<f8'), ('compactness_mean', '<f8'), ('concavity_mean', '<f8'), ('concave_points_mean', '<f8'), ('symmetry_mean', '<f8'), ('fractal_dimension_mean', '<f8'),
('radius_se', '<f8'), ('texture_se', '<f8'), ('perimeter_se', '<f8'), ('area_se', '<f8'), ('smoothness_se', '<f8'), ('compactness_se', '<f8'), ('concavity_se', '<f8'), ('concave_points_se', '<f8'), ('symmetry_se', '<f8'), ('fractal_dimension_se', '<f8'), ('radius_worst', '<f8'), ('texture_worst', '<f8'), ('perimeter_worst', '<f8'), ('area_worst', '<f8'), ('smoothness_worst', '<f8'), ('compactness_worst', '<f8'), ('concavity_worst', '<f8'), ('concave_points_worst', '<f8'), ('symmetry_worst', '<f8'), ('fractal_dimension_worst', '<f8')], names=True)
x = np.column_stack((
cancer_data['radius_mean'], cancer_data['texture_mean'], cancer_data['perimeter_mean'], cancer_data['area_mean'], cancer_data['smoothness_mean'], cancer_data['compactness_mean'], cancer_data['concavity_mean'], cancer_data['concave_points_mean'], cancer_data['symmetry_mean'], cancer_data['fractal_dimension_mean'],
cancer_data['radius_se'], cancer_data['texture_se'], cancer_data['perimeter_se'], cancer_data['area_se'], cancer_data['smoothness_se'], cancer_data['compactness_se'], cancer_data['concavity_se'], cancer_data['concave_points_se'], cancer_data['symmetry_se'], cancer_data['fractal_dimension_se'],
cancer_data['radius_worst'], cancer_data['texture_worst'], cancer_data['perimeter_worst'], cancer_data['area_worst'], cancer_data['smoothness_worst'], cancer_data['compactness_worst'], cancer_data['concavity_worst'], cancer_data['concave_points_worst'], cancer_data['symmetry_worst'], cancer_data['fractal_dimension_worst']
))
x = x / np.max(x, axis=0)
y = np.array([float(item==b'B') for item in cancer_data['diagnosis']]).reshape(len(x), 1)
# Grab input from the user and print it to stdout
population_count = int(input('population_count: '))
print('{}'.format(population_count))
population_size = int(input('poplation_size: '))
print('{}'.format(population_size))
node_cap = int(input('node_cap: '))
print('{}'.format(node_cap))
generations = int(input('generations: '))
print('{}'.format(generations))
target_accuracy = float(input('target_accuracy: '))
print('{}%'.format(target_accuracy*100))
r = int(input("random_seed: "))
print(str(r))
train_labels = utils.jit_to_categorical(
mnist_io.labels_from_file(
os.path.join(dataset,
"train-labels-idx1-ubyte\\train-labels.idx1-ubyte"),
60000), 10)
test_images = mnist_io.images_from_file(
os.path.join(dataset, "t10k-images-idx3-ubyte\\t10k-images.idx3-ubyte"),
10000)
test_images = test_images.reshape(10000, 784).astype('float32')
test_images /= 255
test_labels = utils.jit_to_categorical(
mnist_io.labels_from_file(
os.path.join(dataset, "t10k-labels-idx1-ubyte/t10k-labels.idx1-ubyte"),
10000), 10)
with utils.OutSplit('mnist_evolution'):
evolved_network = evolve_node_count(train_images,
train_labels,
test_images,
test_labels,
utils.jit_categorical_compare,
population_count=10,
population_size=15,
node_cap=1500,
generations=50,
target_accuracy=0.95,
r=4)
print(evolved_network.connections, "\n", evolved_network.weights, "\n",
evolved_network.learning_rate)
import evolutionary_module as evm
import numpy as np
import mnist_io
import utils
r = 23
np.random.seed(r)
with utils.OutSplit('multiple_mnist'):
x = mnist_io.images_from_file(
'datasets\\train-images-idx3-ubyte\\train-images.idx3-ubyte').reshape(
60000, 784).astype('float64') / 255
y = utils.jit_to_categorical(
mnist_io.labels_from_file(
'datasets\\train-labels-idx1-ubyte\\train-labels.idx1-ubyte'), 10)
val_x = mnist_io.images_from_file(
'datasets\\t10k-images-idx3-ubyte\\t10k-images.idx3-ubyte').reshape(
10000, 784).astype('float64') / 255
val_y = utils.jit_to_categorical(
mnist_io.labels_from_file(
'datasets\\t10k-labels-idx1-ubyte\\t10k-labels.idx1-ubyte'), 10)
# Grab input from the user and print it to stdout
population_count = int(input('population_count: '))
print(population_count)
population_size = int(input('poplation_size: '))
print(population_size)
node_cap = int(input('node_cap: '))
print(node_cap)
generations = int(input('generations: '))
print(generations)