from dlgo.nn import load_mnist
from dlgo.nn import network
from dlgo.nn.layers import DenseLayer, ActivationLayer
# first load training and test data
training_data, test_data = load_mnist.load_data()
# Next initialize a sequential neural network
net = network.SequentialNetwork()
# you can then add dense and activation layers one by one
net.add(DenseLayer(784, 392))
net.add(ActivationLayer(392))
net.add(DenseLayer(392, 196))
net.add(ActivationLayer(196))
net.add(DenseLayer(196, 10))
net.add(ActivationLayer(10))
# the final layer has size 10, the number of classes to predict
# You can now easily train the model by specifying train and test data, the number of epochs, the mini-batch size and the learning rate.
net.train(training_data,
epochs=10,
mini_batch_size=10,
learning_rate=3.0,
test_data=test_data)
from dlgo.nn import load_mnist
from dlgo.nn import network
from dlgo.nn.layers import DenseLayer, ActivationLayer
training_data, test_data = load_mnist.load_data(
) # First, load training and test data.
net = network.SequentialNetwork(
) # Next, initialize a sequential neural network.
net.add(DenseLayer(
784, 392)) # You can then add dense and activation layers one by one.
net.add(ActivationLayer(392))
net.add(DenseLayer(392, 196))
net.add(ActivationLayer(196))
net.add(DenseLayer(196, 10))
net.add(ActivationLayer(
10)) # The final layer has size 10, the number of classes to predict.
# You can now easily train the model by specifying train and test data,
# the number of epochs, the mini-batch size and the learning rate.
net.train(training_data,
epochs=10,
mini_batch_size=10,
learning_rate=3.0,
test_data=test_data)
return np.average(filtered_array, axis=0)
def predict(x, W, b):
return sigmoid_double(np.dot(W, x) + b)
def evaluate(data, digit, threshold, W, b):
correct_predictions = 0
for x in data:
if (predict(x[0], W, b) > threshold) == (np.argmax(x[1]) == digit):
correct_predictions += 1
return correct_predictions / float(len(data))
train, test = load_data()
W_8 = np.transpose(average_digit(train, 8))
b_8 = -45
# print(predict(train[2][0], W_8, b_8))
# print(predict(train[17][0], W_8, b_8))
print(evaluate(data=train, digit=8, threshold=0.5, W=W_8, b=b_8))
print(evaluate(data=test, digit=8, threshold=0.5, W=W_8, b=b_8))
print(
evaluate(data=[x for x in test if np.argmax(x[1]) == 8],
digit=8,
threshold=0.5,
W=W_8,
b=b_8))
