def train(train_x, train_y, layers):
nn = NeuralNetwork(layer_dims=layers,
keep_prop=0.8,
weight_factor=len(layers) > 2 and 'deep' or 0.01)
def print_cost(iteration_num, cost):
if iteration_num % 100 == 0:
print("Cost after iteration %d: %f" % (iteration_num, cost))
nn.train(train_x, train_y, num_iterations=2500, callback=print_cost)
return nn
def train(train_x, train_y, test_x, test_y):
def print_cost(iteration_num, cost):
if iteration_num % 100 == 0:
print("Cost after iteration %d: %f" % (iteration_num, cost))
nn = NeuralNetwork(layer_dims=[20, 3, 1],
num_iterations=3000,
weight_factors='deep',
callback=print_cost)
nn.train(train_x, train_y)
p = nn.predict(train_x)
print('Accuracy: %f' % np.mean(p == train_y))
p = nn.predict(test_x)
print('Accuracy: %f' % np.mean(p == test_y))