return (inputTrain, outputTrain, inputTest, outputTest)
def run(dataset, arhitecture, learning_rate, eval_every, stop):
input_size = dataset["train_imgs"][0].shape
nn = FeedForward(input_size, arhitecture)
print(nn.to_string())
return train_nn(nn, dataset, learning_rate, eval_every, stop)
if __name__ == "__main__":
parser = ArgumentParser()
parser.add_argument("--learning_rate", type = float, default = 0.001,
help="Learning rate")
parser.add_argument("--eval_every", type = int, default = 2000,
help="Learning rate")
args = parser.parse_args()
#dataset = load_mnist()
dataset = load_cifrar()
input_size = dataset["train_imgs"][0].shape
nn = FeedForward(input_size, [(CONV, (6, 28, 28), 5, 1), (RELU, -1), (MAX_POOLING, (6, 14, 14)), (CONV, (16, 10, 10), 5, 1), (RELU, -1), (MAX_POOLING, (16, 5, 5)),
(LINEARIZE, -1), (FULLY_CONNECTED, 120), (FULLY_CONNECTED, 84), (FULLY_CONNECTED, 10) ,(SOFTMAX, -1)])
#nn = FeedForward(input_size, [(LINEARIZE, -1), (FULLY_CONNECTED, 300), (TANH, -1), (FULLY_CONNECTED, 100), (TANH, -1), (FULLY_CONNECTED, 10), (SOFTMAX, -1)])
print(nn.to_string())
train_nn(nn, dataset, args.learning_rate, args.eval_every, 10000)
def run(dataset, arhitecture, learning_rate, eval_every, stop): input_size = dataset["train_imgs"][0].shape nn = FeedForward(input_size, arhitecture) print(nn.to_string()) return train_nn(nn, dataset, learning_rate, eval_every, stop)
nn.update_parameters(args.learning_rate)
# Evaluate the network
if cnt % args.eval_every == 0:
test_acc, test_cm = \
eval_nn(nn, data["test_imgs"], data["test_labels"])
train_acc, train_cm = \
eval_nn(nn, data["train_imgs"], data["train_labels"], 5000)
print("Train acc: %2.6f ; Test acc: %2.6f" % (train_acc, test_acc))
pylab.imshow(test_cm)
pylab.draw()
matplotlib.pyplot.pause(0.001)
if __name__ == "__main__":
parser = ArgumentParser()
parser.add_argument("--learning_rate", type = float, default = 0.001,
help="Learning rate")
parser.add_argument("--eval_every", type = int, default = 200,
help="Learning rate")
args = parser.parse_args()
mnist = load_mnist()
input_size = mnist["train_imgs"][0].size
print input_size
nn = FeedForward(input_size, [(300, logistic), (10, identity)])
print(nn.to_string())
train_nn(nn, mnist, args)
