import mnist_loader

import network

import numpy as np

def get_data(filename, num_labels, delimiter=','):

train_data = []

with open(filename) as fin:

for line in fin:

line_arr = np.array(list(map(float, line.strip().split(delimiter))))

x, y = line_arr[:-num_labels], line_arr[-num_labels:]

x = x.reshape(len(x), 1)

y = y.reshape(len(y), 1)

train_data.append((x, y))

return train_data

def train_images():

training_data, validation_data, test_data = mnist_loader.load_data_wrapper()

net = network.Network([784, 30, 10])

net.SGD(training_data, epochs=3, mini_batch_size=5, eta=3.0)

def train_2dim():

training_data = get_data("data/data2.txt", num_labels=1)

net = network.Network([2, 5, 1])

num_epochs, batch_size, learn_rate = 90, 10, 3.0

net.SGD(training_data, num_epochs, batch_size, learn_rate) # works

print("output at 0, 0:" + str(net.feedforward(np.array([0., 0.]).reshape(2, 1))))

print("output at 10, 10:" + str(net.feedforward(np.array([10., 10.]).reshape(2, 1))))

print("num_epochs, batch_size, learn_rate = {0}, {1}, {2}".format(num_epochs, batch_size, learn_rate))

for b, w in zip(net.biases, net.weights):

print(np.concatenate((b, w), 1))

if __name__ == "__main__":

train_2dim()