def play(self):
try:
mkdir('examples/connect4/genomes')
except OSError:
pass
players = input('Players: ')
if players == 0:
output = input('Output: ')
for _ in xrange(input('Iterations: ')):
processes = []
for i in xrange(20):
processes.append(Process(target=self.thread, kwargs={'i': i, 'output': output}))
processes[-1].start()
for process in processes:
process.join()
elif players == 1:
roll = randint(0, 19)
ann = ANN(inputs=49, outputs=3, hidden=49, rows=5)
try:
ann.genome = load(open('examples/connect4/genomes/genome{0:02d}.p'.format(roll), 'rb'))
except IOError:
pass
if roll < 10:
Connect4([ann, None]).play()
else:
Connect4([None, ann]).play()
elif players == 2:
Connect4().play()
def setUp(self):
self.G=ANN([[None]], sum)
self.costs={(0,0):None} # pair: cost dict.
self.heads={0:()}
self.tails={0:()}
self.sources=(0,)
self.sinks=(0,)
self.order=1
# Should I be returning tuples or lists?
self.outputs={(0.1,):{0:0.1},(1,):{0:1},(5,):{0:5}}
class TestANNBase(TestGraphBase):
def setUp(self):
self.G=ANN([[None]], sum)
self.costs={(0,0):None} # pair: cost dict.
self.heads={0:()}
self.tails={0:()}
self.sources=(0,)
self.sinks=(0,)
self.order=1
# Should I be returning tuples or lists?
self.outputs={(0.1,):{0:0.1},(1,):{0:1},(5,):{0:5}}
def testFeedforward(self):
for inputs, output in self.outputs.items():
self.assertEqual({i:round(o, 6) for i, o in
self.G.feedforward(inputs).items()},output)
import numpy as np
import pandas as pd
filename = 'data_resource_usage_5Minutes_6176858948.csv'
f = '5m'
foldername = 'data/'
# parameters
idx = (6336, 6336 + 1584)
sliding_windows = [2, 3, 5]
pop_sizes = [100, 150, 200]
cross_rates = [0.7, 0.8, 0.9]
mutate_rates = [0.01, 0.02, 0.03]
method_statistic = [0, 1, 2]
n_expanded = [2, 3, 4]
activations = [3, 2, 1, 0]
df = pd.read_csv(foldername + filename,
header=None,
index_col=False,
usecols=[3, 4])
df.dropna(inplace=True)
dataset_original = df.values
for sliding in sliding_windows:
for ms in method_statistic:
for activation in activations:
p = ANN(dataset_original, idx[0], idx[1], sliding, ms, activation,
f)
p.train(epochs=1000)
