Python implementation of the convolutional neural network SimpleConvNet
# coding: utf-8
import sys, os
sys.path.append('D:\AI learning source code') # 为了导入父目录的文件而进行的设定
import pickle
import numpy as np
from collections import OrderedDict
from common.layers import *
from common.gradient import numerical_gradient
class SimpleConvNet:
"""简单的ConvNet
conv - relu - pool - affine - relu - affine - softmax
Parameters
----------
input_size : 输入大小（MNIST的情况下为784）
hidden_size_list : 隐藏层的神经元数量的列表（e.g. [100, 100, 100]）
output_size : 输出大小（MNIST的情况下为10）
activation : 'relu' or 'sigmoid'
weight_init_std : 指定权重的标准差（e.g. 0.01）
指定'relu'或'he'的情况下设定“He的初始值”
指定'sigmoid'或'xavier'的情况下设定“Xavier的初始值”
"""
def __init__(self, input_dim=(1, 28, 28),
conv_param={
'filter_num':30, 'filter_size':5, 'pad':0, 'stride':1},
hidden_size=100, output_size=10, weight_init_std=0.01):
filter_num = conv_param['filter_num']
filter_size = conv_param['filter_size']
filter_pad = conv_param['pad']
filter_stride = conv_param['stride']
input_size = input_dim[1]
conv_output_size = (input_size - filter_size + 2*filter_pad) / filter_stride + 1
pool_output_size = int(filter_num * (conv_output_size/2) * (conv_output_size/2))
# 初始化权重
self.params = {
}
self.params['W1'] = weight_init_std * \
np.random.randn(filter_num, input_dim[0], filter_size, filter_size)
self.params['b1'] = np.zeros(filter_num)
self.params['W2'] = weight_init_std * \
np.random.randn(pool_output_size, hidden_size)
self.params['b2'] = np.zeros(hidden_size)
self.params['W3'] = weight_init_std * \
np.random.randn(hidden_size, output_size)
self.params['b3'] = np.zeros(output_size)
# 生成层
self.layers = OrderedDict()
self.layers['Conv1'] = Convolution(self.params['W1'], self.params['b1'],
conv_param['stride'], conv_param['pad'])
self.layers['Relu1'] = Relu()
self.layers['Pool1'] = Pooling(pool_h=2, pool_w=2, stride=2)
self.layers['Affine1'] = Affine(self.params['W2'], self.params['b2'])
self.layers['Relu2'] = Relu()
self.layers['Affine2'] = Affine(self.params['W3'], self.params['b3'])
self.last_layer = SoftmaxWithLoss()
def predict(self, x):
for layer in self.layers.values():
x = layer.forward(x)
return x
def loss(self, x, t):
"""求损失函数
参数x是输入数据、t是教师标签
"""
y = self.predict(x)
return self.last_layer.forward(y, t)
def accuracy(self, x, t, batch_size=100):
if t.ndim != 1 : t = np.argmax(t, axis=1)
acc = 0.0
for i in range(int(x.shape[0] / batch_size)):
tx = x[i*batch_size:(i+1)*batch_size]
tt = t[i*batch_size:(i+1)*batch_size]
y = self.predict(tx)
y = np.argmax(y, axis=1)
acc += np.sum(y == tt)
return acc / x.shape[0]
def numerical_gradient(self, x, t):
"""求梯度（数值微分）
Parameters
----------
x : 输入数据
t : 教师标签
Returns
-------
具有各层的梯度的字典变量
grads['W1']、grads['W2']、...是各层的权重
grads['b1']、grads['b2']、...是各层的偏置
"""
loss_w = lambda w: self.loss(x, t)
grads = {
}
for idx in (1, 2, 3):
grads['W' + str(idx)] = numerical_gradient(loss_w, self.params['W' + str(idx)])
grads['b' + str(idx)] = numerical_gradient(loss_w, self.params['b' + str(idx)])
return grads
def gradient(self, x, t):
"""求梯度（误差反向传播法）
Parameters
----------
x : 输入数据
t : 教师标签
Returns
-------
具有各层的梯度的字典变量
grads['W1']、grads['W2']、...是各层的权重
grads['b1']、grads['b2']、...是各层的偏置
"""
# forward
self.loss(x, t)
# backward
dout = 1
dout = self.last_layer.backward(dout)
layers = list(self.layers.values())
layers.reverse()
for layer in layers:
dout = layer.backward(dout)
# 设定
grads = {
}
grads['W1'], grads['b1'] = self.layers['Conv1'].dW, self.layers['Conv1'].db
grads['W2'], grads['b2'] = self.layers['Affine1'].dW, self.layers['Affine1'].db
grads['W3'], grads['b3'] = self.layers['Affine2'].dW, self.layers['Affine2'].db
return grads
def save_params(self, file_name="params.pkl"):
params = {
}
for key, val in self.params.items():
params[key] = val
with open(file_name, 'wb') as f:
pickle.dump(params, f)
def load_params(self, file_name="params.pkl"):
with open(file_name, 'rb') as f:
params = pickle.load(f)
for key, val in params.items():
self.params[key] = val
for i, key in enumerate(['Conv1', 'Affine1', 'Affine2']):
self.layers[key].W = self.params['W' + str(i+1)]
self.layers[key].b = self.params['b' + str(i+1)]
1
def __init__(self, input_dim=(1, 28, 28),
conv_param={
'filter_num':30, 'filter_size':5, 'pad':0, 'stride':1},
hidden_size=100, output_size=10, weight_init_std=0.01):
filter_num = conv_param['filter_num']
filter_size = conv_param['filter_size']
filter_pad = conv_param['pad']
filter_stride = conv_param['stride']
input_size = input_dim[1]
conv_output_size = (input_size - filter_size + 2*filter_pad) / filter_stride + 1
pool_output_size = int(filter_num * (conv_output_size/2) * (conv_output_size/2))
In the construction method, we set the parameters of the convolutional neural network, the number of input neurons, filter_num (number of convolution kernels), filter_size (convolution kernel size), pad (filling amount), stride (convolution kernel stride) , the number of neurons in the hidden layer, the number of output neurons, and the standard deviation of the initial weight values.
According to the parameters passed in, the program calculates the output size of the convolution layer and pooling layer.
2
# 初始化权重
self.params = {
}
self.params['W1'] = weight_init_std * \
np.random.randn(filter_num, input_dim[0], filter_size, filter_size)
self.params['b1'] = np.zeros(filter_num)
self.params['W2'] = weight_init_std * \
np.random.randn(pool_output_size, hidden_size)
self.params['b2'] = np.zeros(hidden_size)
self.params['W3'] = weight_init_std * \
np.random.randn(hidden_size, output_size)
self.params['b3'] = np.zeros(output_size)
Initialize the weights, save the values of each weight and bias as a dictionary and assign each weight to a random normal * weight_init_std, and set the initial value of the bias to 0
3
# 生成层
self.layers = OrderedDict()
self.layers['Conv1'] = Convolution(self.params['W1'], self.params['b1'],
conv_param['stride'], conv_param['pad'])
self.layers['Relu1'] = Relu()
self.layers['Pool1'] = Pooling(pool_h=2, pool_w=2, stride=2)
self.layers['Affine1'] = Affine(self.params['W2'], self.params['b2'])
self.layers['Relu2'] = Relu()
self.layers['Affine2'] = Affine(self.params['W3'], self.params['b3'])
self.last_layer = SoftmaxWithLoss()
Build the network structure and save the network as an OrderedDict.
The structure of the network:
convolutional layer – ReLU – pooling – Affine (summation) – ReLU – Affine – softmax (output)
3
def gradient(self, x, t):
"""求梯度（误差反向传播法）
Parameters
----------
x : 输入数据
t : 教师标签
Returns
-------
具有各层的梯度的字典变量
grads['W1']、grads['W2']、...是各层的权重
grads['b1']、grads['b2']、...是各层的偏置
"""
# forward
self.loss(x, t)
# backward
dout = 1
dout = self.last_layer.backward(dout)
layers = list(self.layers.values())
layers.reverse()
for layer in layers:
dout = layer.backward(dout)
# 设定
grads = {
}
grads['W1'], grads['b1'] = self.layers['Conv1'].dW, self.layers['Conv1'].db
grads['W2'], grads['b2'] = self.layers['Affine1'].dW, self.layers['Affine1'].db
grads['W3'], grads['b3'] = self.layers['Affine2'].dW, self.layers['Affine2'].db
return grads
Use the backpropagation gradient descent method to find the gradient of each weight and bias, and its implementation process is consistent with that of the fully connected network.
training network
# coding: utf-8
import sys, os
sys.path.append('D:\AI learning source code') # 为了导入父目录的文件而进行的设定
import numpy as np
import matplotlib.pyplot as plt
from dataset.mnist import load_mnist
from simple_convnet import SimpleConvNet
from common.trainer import Trainer
# 读入数据
(x_train, t_train), (x_test, t_test) = load_mnist(flatten=False)
# 处理花费时间较长的情况下减少数据
x_train, t_train = x_train[:5000], t_train[:5000]
#x_test, t_test = x_test[:1000], t_test[:1000]
max_epochs = 20
network = SimpleConvNet(input_dim=(1,28,28),
conv_param = {
'filter_num': 30, 'filter_size': 5, 'pad': 0, 'stride': 1},
hidden_size=100, output_size=10, weight_init_std=0.01)
trainer = Trainer(network, x_train, t_train, x_test, t_test,
epochs=max_epochs, mini_batch_size=100,
optimizer='Adam', optimizer_param={
'lr': 0.001},
evaluate_sample_num_per_epoch=1000)
trainer.train()
# 保存参数
network.save_params("params.pkl")
print("Saved Network Parameters!")
# 绘制图形
markers = {
'train': 'o', 'test': 's'}
x = np.arange(max_epochs)
plt.plot(x, trainer.train_acc_list, marker='o', label='train', markevery=2)
plt.plot(x, trainer.test_acc_list, marker='s', label='test', markevery=2)
plt.xlabel("epochs")
plt.ylabel("accuracy")
plt.ylim(0, 1.0)
plt.legend(loc='lower right')
plt.show()
The prediction accuracy image obtained after running 5000 times of training:
It can be seen that the accuracy of the convolutional neural network for image recognition problems is much stronger than that of the ordinary fully connected network. In the end, it can reach almost 98% prediction accuracy
CNN visualization
In the following program, we use 30 5 X 5 convolution kernels with a channel of 1 to check the grayscale image, and visualize the weights of each convolution kernel before and after learning
# coding: utf-8
import numpy as np
import matplotlib.pyplot as plt
from simple_convnet import SimpleConvNet
def filter_show(filters, nx=8, margin=3, scale=10):
"""
c.f. https://gist.github.com/aidiary/07d530d5e08011832b12#file-draw_weight-py
"""
FN, C, FH, FW = filters.shape
ny = int(np.ceil(FN / nx))
fig = plt.figure()
fig.subplots_adjust(left=0, right=1, bottom=0, top=1, hspace=0.05, wspace=0.05)
for i in range(FN):
ax = fig.add_subplot(ny, nx, i+1, xticks=[], yticks=[])
ax.imshow(filters[i, 0], cmap=plt.cm.gray_r, interpolation='nearest')
plt.show()
network = SimpleConvNet()
# 随机进行初始化后的权重
filter_show(network.params['W1'])
# 学习后的权重
network.load_params("params.pkl")
filter_show(network.params['W1'])
Before learning:
After learning:
It can be seen that before learning, the initial value of the weight is randomly generated, and the black and white areas of the convolution kernel are randomly separated. But after learning, you can see that the black areas present specific patterns (called blobs).
Different patterns represent the features of image edges and blobs. A convolution kernel with a horizontal pattern will better extract horizontal edges in the image, a convolution kernel with a vertical pattern will extract vertical edges, and so on.
In deep learning, the subsequent CNN network will perform further feature extraction on the basic pattern features obtained by the first layer of CNN, and the extracted features will gradually change from local to overall.
For example, for the AlexNet of the 8th layer in the above figure. The first layer of neurons will extract edges and plaques, the third layer will extract object textures, the fifth layer will extract object parts, and the last fully connected layer will classify objects according to the characteristics of object parts
Representative CNN
1 LeNet
LeNet is the earliest convolutional neural network, and its structure has formed the prototype of the convolution-pooling structure. However, unlike the commonly used CNN model now, LeNet uses sigmoid as the activation function (reLU is often used now), and LeNet uses subsampling to reduce the size of intermediate data, and now CNN generally uses Max Pooling
Compared with LeNet, AlexNet has the same basic network structure, but has the following differences :
Use ReLU as activation function
Use locally regularized LRN layer
Use Dropout