设计思路
设计代码
import mnist_loader
from network3 import Network
from network3 import ConvPoolLayer, FullyConnectedLayer, SoftmaxLayer
training_data, validation_data, test_data = mnist_loader.load_data_wrapper()
mini_batch_size = 10
#NN算法：sigmoid函数；准确率97%
net = Network([
FullyConnectedLayer(n_in=784, n_out=100),
SoftmaxLayer(n_in=100, n_out=10)], mini_batch_size)
net.SGD(training_data, 60, mini_batch_size, 0.1, validation_data, test_data)
#CNN算法：1层Convolution+sigmoid函数；准确率98.78%
net = Network([
ConvPoolLayer(image_shape=(mini_batch_size, 1, 28, 28),
filter_shape=(20, 1, 5, 5),
poolsize=(2, 2)),
FullyConnectedLayer(n_in=20*12*12, n_out=100),
SoftmaxLayer(n_in=100, n_out=10)], mini_batch_size)
#CNN算法：2层Convolution+sigmoid函数；准确率99.06%。层数过多并不会使准确率大幅度提高，有可能overfit，合适的层数需要通过实验验证出来，并不是越多越好
net = Network([
ConvPoolLayer(image_shape=(mini_batch_size, 1, 28, 28),
filter_shape=(20, 1, 5, 5),
poolsize=(2, 2)),
ConvPoolLayer(image_shape=(mini_batch_size, 20, 12, 12),
filter_shape=(40, 20, 5, 5),
poolsize=(2, 2)),
FullyConnectedLayer(n_in=40*4*4, n_out=100),
SoftmaxLayer(n_in=100, n_out=10)], mini_batch_size)
#CNN算法：用Rectified Linear Units即f(z) = max(0, z)，代替sigmoid函数；准确率99.23%
net = Network([
ConvPoolLayer(image_shape=(mini_batch_size, 1, 28, 28),
filter_shape=(20, 1, 5, 5),
poolsize=(2, 2),
activation_fn=ReLU), #激活函数采用ReLU函数
ConvPoolLayer(image_shape=(mini_batch_size, 20, 12, 12),
filter_shape=(40, 20, 5, 5),
poolsize=(2, 2),
activation_fn=ReLU),
FullyConnectedLayer(n_in=40*4*4, n_out=100, activation_fn=ReLU),
SoftmaxLayer(n_in=100, n_out=10)], mini_batch_size)
#CNN算法：用ReLU函数+增大训练集25万(原先50000*5，只需将每个像素向上下左右移动一个像素)；准确率99.37%
$ python expand_mnist.py #将图片像素向上下左右移动
expanded_training_data, _, _ = network3.load_data_shared("../data/mnist_expanded.pkl.gz")
net = Network([
ConvPoolLayer(image_shape=(mini_batch_size, 1, 28, 28),
filter_shape=(20, 1, 5, 5),
poolsize=(2, 2),
activation_fn=ReLU),
ConvPoolLayer(image_shape=(mini_batch_size, 20, 12, 12),
filter_shape=(40, 20, 5, 5),
poolsize=(2, 2),
activation_fn=ReLU),
FullyConnectedLayer(n_in=40*4*4, n_out=100, activation_fn=ReLU),
SoftmaxLayer(n_in=100, n_out=10)], mini_batch_size)
net.SGD(expanded_training_data, 60, mini_batch_size, 0.03,validation_data, test_data, lmbda=0.1)
#CNN算法：用ReLU函数+增大训练集25万+dropout(随机选取一半神经元)用在最后的FullyConnected层；准确率99.60%
expanded_training_data, _, _ = network3.load_data_shared("../data/mnist_expanded.pkl.gz")
net = Network([
ConvPoolLayer(image_shape=(mini_batch_size, 1, 28, 28),
filter_shape=(20, 1, 5, 5),
poolsize=(2, 2),
activation_fn=ReLU),
ConvPoolLayer(image_shape=(mini_batch_size, 20, 12, 12),
filter_shape=(40, 20, 5, 5),
poolsize=(2, 2),
activation_fn=ReLU),
FullyConnectedLayer(
n_in=40*4*4, n_out=1000, activation_fn=ReLU, p_dropout=0.5),
FullyConnectedLayer(
n_in=1000, n_out=1000, activation_fn=ReLU, p_dropout=0.5),
SoftmaxLayer(n_in=1000, n_out=10, p_dropout=0.5)],
mini_batch_size)
net.SGD(expanded_training_data, 40, mini_batch_size, 0.03,validation_data, test_data)