def crossover(parents):
parent1, parent2 = parents
child_network = []
for i in range(len(parent1.network)):
temp_layer = parent1.network[i].copy()
for j in range(temp_layer.shape[0]):
temp_layer[j, :] = random.choice(
[parent1.network[i][j, :], parent2.network[i][j, :]])
child_network.append(temp_layer)
child = nn.neuralNet(0, 0)
child.network = child_network
return child
import numpy
from neuralNet import neuralNet
# set variables and create neural network object
input_nodes = 784
hidden_nodes = 100
output_nodes = 10
learning_rate = .3
net = neuralNet(input_nodes, hidden_nodes, output_nodes, learning_rate)
# train neural network
training_data_file = open("mnist_dataset/mnist_train.csv", 'r')
for line in training_data_file:
all_values = line.split(',')
inputs = (numpy.asfarray(all_values[1:]) / 255.0 * 0.99) + 0.01
targets = numpy.zeros(output_nodes) + 0.01
targets[int(all_values[0])] = 0.99
net.train(inputs, targets)
pass
training_data_file.close()
# test neural network
testing_data_file = open("mnist_dataset/mnist_test.csv", 'r')
test_results = []
for line in testing_data_file:
all_values = line.split(',')
test_results.append([
if line:
line = [float(i) for i in line]
inputs.append(line)
with open('output.txt') as f:
outputs = []
for line in f:
line = line.split()
if line:
line = [int(i) for i in line]
outputs.append(line)
input = np.array(inputs)
output = np.array(outputs)
nn = neuralNet(400,30,10)
# Training
# ---
# Batch training
nn.trainBatch(input,output,20)
# Live training
#tests = np.size(input,0)
#acc = 0
#for i in xrange(0, tests):
# if (np.argmax(nn.trainLive(input[[i],:],output[i,0])) == output[i,0]):
# acc = acc + 1
#acc = acc / float(tests) * 100
#print("Live training accuracy: %f" % (acc))
import neuralNet as neural
myNeuralNetwork = None
netShape = []
netShape.append(2)
netShape.append(5)
netShape.append(5)
netShape.append(2)
myNeuralNetwork = neural.neuralNet(netShape)
myNeuralNetwork.visualise()
fileHandle = None
fileData = None
trainingList = None
fileHandle = open( 'trainingData.txt', 'r' )
fileData = fileHandle.read()
fileHandle.close()
trainingList = fileData.split(',')
fileHandle1 = None
fileData1 = None
trainingLists = None
fileHandle1 = open('trainingTarget.txt', 'r' )
fileData1 = fileHandle1.read()
fileHandle1.close()
trainingLists = fileData1.split(',')
tl = []
for item in trainingLists:
screen.fill(background_col)
screen.blit(temp_ground[0], temp_ground_rect)
if temp_dino.isBlinking:
screen.blit(logo, logo_rect)
screen.blit(callout, callout_rect)
temp_dino.draw()
pygame.display.update()
clock.tick(FPS)
if temp_dino.isJumping == False and temp_dino.isBlinking == False:
gameStart = True
# Variáveis de jogo
redes = [nn.neuralNet(3, 1, [5]) for i in range(n_rex)]
dead = []
gamespeed = 12
gameOver = [False for n in range(n_rex)]
p_gameOver = False
playerDino = [Dino(44, 47) for n in range(n_rex)]
player = (Dino(44, 47, True))
keys = [non for n in range(n_rex)]
cacti = pygame.sprite.Group()
pteras = pygame.sprite.Group()
clouds = pygame.sprite.Group()
last_obstacle = pygame.sprite.Group()
dists = []
heights = []
#for i in range(0,w):
# for j in range(0,h):
# r,g,b = [myTestingImage[i,j,0]/255.0,myTestingImage[i,j,1]/255.0,myTestingImage[i,j,2]/255.0]
# myImage[i,j] = [r,g,b]
myTestingImages.append(testing[k])
myTestingLabels.append(myTestingLabel[0])
print("Data loading completed successfully")
#Before batch training, we validate some important things
myFakeImage = np.array(([[[0.8,0.2,1.0]]]))
myFakeImage = myFakeImage.flatten()
myNN = nn.neuralNet(randomWeightsRange=randomWeightsRange,alpha=alpha,mi=mi,batchSize=batchSize)
myNN.activateNetwork(myFakeImage)
# myNN.transparent()
newWeights,Loss = myNN.batchUpdate([myFakeImage],[1])
myNN.activateNetwork(myFakeImage)
# myNN.transparent()
exit()
# Verify improvements on the testing set
sumC = 0
sum = 0
for i in range(0,int(testing.size/3072)):
myImageFlattened = (testing[i]).flatten()