转载请注明出处。谢谢。
本博文根据 coursera 吴恩达 深度学习整理。作为理解神经网络的基础。
本章主要内容分为两个部分,第一理解浅层网络中的前向和后向传播规则,尤其要注意其中的标注规范;第二了解到实现的细节,包括激活函数的选择以及权重初始化等更为工程化的内容。
一、知识点
1. 单个隐藏层的神经网络
a) 命名
第一层:输入层; 最后一层:输出层;中间层:隐藏层
对于标记, 上标 i 表示层数,下标 j 表示神经元的个数
对于输入样本X,, 记做
, 上标 i 表示第 i 个样本,下标 j 表示第 j 个特征。
b) 向量化前向计算
和逻辑回归类似,可以写出以上公式。利用下图( 单个神经元)可以增进理解单个神经元的作用和一层神经元的作用,任意一个神经元都是由一个线性变换和一个激活函数组合而成,该层的输出为下一层的输入:
有两个地方值得注意:
首先矩阵W,在逻辑回归中是w_T,在这里W其实是每一个神经元的w_t的组合,统一记做了W,即
其次,中间的函数 g 表示了任意激活函数,可以是sigmoid,tanh, ReLU, Leaky ReLU 中任意一个。但是由于这是二分类问题,所以最后的激活函数仍旧是sigmoid。
c) 向量化的反向传播
最简便的方法就是依据链式法则和前向公式直接计算。第二层的反向传播和逻辑回归类似。但第一层有细微差别:
dz2 = A2 - Y
dw2 = 1.0/m * np.dot(dz2, A1.T)
db2 = 1.0/m * np.sum(dz2, axis=1, keepdims=True)
dz1 = np.dot(w2.T, dz2) * g_(Z1)
dw1 = 1.0/m * np.dot(dz1,X.T)
db1 = 1.0/m * np.sum(dz1,axis=1,keepdims=True)
2. 激活函数
(1) sigmoid:
def sigmoid(x):
return 1.0 / (1.0 + np.exp(-x))
优点:输出是在(0,1)开区间内,容易联想到概率。
缺点:当输入极大或极小,函数的梯度就变得很小了,几乎为零。在反向传播的过程中,容易造成梯度饱和(梯度弥散);函数输出不是以0为中心的,这样会使权重更新效率降低;sigmod函数要进行指数运算,相对比较慢
(2) tanh :
def tanh(x):
return 2 * sigmoid(2 * x) - 1
tanh是双曲正切函数,tanh函数和sigmod函数的曲线是比较相近的。同样不利于梯度下降。但其在整个函数是以0为中心的,因而特性相对好一些。一般二分类问题中,隐藏层用tanh函数,输出层用sigmod函数。
(3) ReLU :
def ReLU(x):
return np.maximum(0, x)
该函数能较好解决梯度下降的问题。计算速度要快很多,不需要进行指数运算。但当输入是负数的时候,ReLU是完全不被激活的,在反向传播过程中,输入负数,梯度就会完全恒为零,无法更新。同样它也不是对称的图像。
(4)Leaky ReLu :
def LeakyReLu(x):
return np.maximum(0.01*x, x)
在负数区域内是线性运算,斜率很小,但是不会趋于0.
3. 随机初始化
在多层网络中,不能使用0为初始化权重,因为这会导致同一层的梯度更新为同一个值。最好采用随机初始化:
w = np.random.rand((2,2))*0.01
b = np.zeros((2,1))
### 概念辨析 ###:
Logistic回归的权重w应该随机初始化,而不是全零,因为如果初始化为全零,那么逻辑回归将无法学习到有用的决策边界,因为它将无法“破坏对称性”,是正确的吗?
Logistic回归没有隐藏层。 如果将权重初始化为零,则Logistic回归中的第一个示例x将输出零,但Logistic回归的导数取决于不是零的输入x(因为没有隐藏层)。 因此,在第二次迭代中,如果x不是常量向量,则权值遵循x的分布并且彼此不同。
二、实践
The general methodology to build a Neural Network is to:
1. Define the neural network structure ( # of input units, # of hidden units, etc). 2. Initialize the model's parameters 3. Loop: - Implement forward propagation - Compute loss - Implement backward propagation to get the gradients - Update parameters (gradient descent)
You often build helper functions to compute steps 1-3 and then merge them into one function we call
nn_model()
. Once you've builtnn_model()
and learnt the right parameters, you can make predictions on new data.
1. 定义网络的结构:输入层的大小(特征)、输出层大小(特征)包括隐藏层神经元个数
- n_x: the size of the input layer
- n_h: the size of the hidden layer (set this to 4)
- n_y: the size of the output layer
def layer_sizes(X, Y):
"""
Arguments:
X -- input dataset of shape (input size, number of examples)
Y -- labels of shape (output size, number of examples)
Returns:
n_x -- the size of the input layer
n_h -- the size of the hidden layer
n_y -- the size of the output layer
"""
### START CODE HERE ### (≈ 3 lines of code)
n_x = X.shape[0] # size of input layer
n_h = 4
n_y = Y.shape[0] # size of output layer
### END CODE HERE ###
return (n_x, n_h, n_y)
2. 初始化网络参数
np.random.randn(a,b) * 0.01
对尺寸为 (a,b) 的权重矩阵进行初始化。
np.zeros((a,b))
对尺寸为(a,b)的bias矩阵进行初始化。
# GRADED FUNCTION: initialize_parameters
def initialize_parameters(n_x, n_h, n_y):
"""
Argument:
n_x -- size of the input layer
n_h -- size of the hidden layer
n_y -- size of the output layer
Returns:
params -- python dictionary containing your parameters:
W1 -- weight matrix of shape (n_h, n_x)
b1 -- bias vector of shape (n_h, 1)
W2 -- weight matrix of shape (n_y, n_h)
b2 -- bias vector of shape (n_y, 1)
"""
np.random.seed(2) # we set up a seed so that your output matches ours although the initialization is random.
### START CODE HERE ### (≈ 4 lines of code)
W1 = np.random.randn(n_h, n_x) * 0.01
b1 = np.zeros((n_h, 1))
W2 = np.random.randn(n_y, n_h) * 0.01
b2 = np.zeros((n_y, 1))
### END CODE HERE ###
assert (W1.shape == (n_h, n_x))
assert (b1.shape == (n_h, 1))
assert (W2.shape == (n_y, n_h))
assert (b2.shape == (n_y, 1))
parameters = {"W1": W1,
"b1": b1,
"W2": W2,
"b2": b2}
return parameters
3. 开始循环(根据公式)
3.1 前向传播
# GRADED FUNCTION: forward_propagation
def forward_propagation(X, parameters):
"""
Argument:
X -- input data of size (n_x, m)
parameters -- python dictionary containing your parameters (output of initialization function)
Returns:
A2 -- The sigmoid output of the second activation
cache -- a dictionary containing "Z1", "A1", "Z2" and "A2"
"""
# Retrieve each parameter from the dictionary "parameters"
### START CODE HERE ### (≈ 4 lines of code)
W1 = parameters["W1"]
b1 = parameters["b1"]
W2 = parameters["W2"]
b2 = parameters["b2"]
### END CODE HERE ###
# Implement Forward Propagation to calculate A2 (probabilities)
### START CODE HERE ### (≈ 4 lines of code)
Z1 = np.dot(W1,X)+b1
A1 = np.tanh(Z1)
Z2 = np.dot(W2,A1)+b2
A2 = sigmoid(Z2)
### END CODE HERE ###
assert(A2.shape == (1, X.shape[1]))
cache = {"Z1": Z1,
"A1": A1,
"Z2": Z2,
"A2": A2}
return A2, cache
3.2 计算损失
logprobs = np.multiply(np.log(A2),Y)
cost = - np.sum(logprobs)
# GRADED FUNCTION: compute_cost
def compute_cost(A2, Y, parameters):
"""
Computes the cross-entropy cost given in equation (13)
Arguments:
A2 -- The sigmoid output of the second activation, of shape (1, number of examples)
Y -- "true" labels vector of shape (1, number of examples)
parameters -- python dictionary containing your parameters W1, b1, W2 and b2
Returns:
cost -- cross-entropy cost given equation (13)
"""
m = Y.shape[1] # number of example
# Compute the cross-entropy cost
### START CODE HERE ### (≈ 2 lines of code)
logprobs = np.multiply(np.log(A2),Y)+ np.multiply(np.log(1-A2),1-Y)
cost = -1.0/m* np.sum(logprobs)
### END CODE HERE ###
cost = np.squeeze(cost) # makes sure cost is the dimension we expect.
# E.g., turns [[17]] into 17
assert(isinstance(cost, float))
return cost
3.3 根据公式计算反向传播
# GRADED FUNCTION: backward_propagation
def backward_propagation(parameters, cache, X, Y):
"""
Implement the backward propagation using the instructions above.
Arguments:
parameters -- python dictionary containing our parameters
cache -- a dictionary containing "Z1", "A1", "Z2" and "A2".
X -- input data of shape (2, number of examples)
Y -- "true" labels vector of shape (1, number of examples)
Returns:
grads -- python dictionary containing your gradients with respect to different parameters
"""
m = X.shape[1]
# First, retrieve W1 and W2 from the dictionary "parameters".
### START CODE HERE ### (≈ 2 lines of code)
W1 = parameters["W1"]
W2 = parameters["W2"]
### END CODE HERE ###
# Retrieve also A1 and A2 from dictionary "cache".
### START CODE HERE ### (≈ 2 lines of code)
A1 = cache["A1"]
A2 = cache["A2"]
### END CODE HERE ###
# Backward propagation: calculate dW1, db1, dW2, db2.
### START CODE HERE ### (≈ 6 lines of code, corresponding to 6 equations on slide above)
dZ2 = A2 - Y
dW2 = 1.0/m * np.dot(dZ2,A1.T)
db2 = 1.0/m * np.sum(dZ2,axis=1,keepdims=True)
dZ1 = np.dot(W2.T,dZ2)*(1 - np.power(A1, 2))
dW1 = 1.0/m * np.dot(dZ1,X.T)
db1 = 1.0/m * np.sum(dZ1,axis=1,keepdims=True)
### END CODE HERE ###
grads = {"dW1": dW1,
"db1": db1,
"dW2": dW2,
"db2": db2}
return grads
3.4 参数更新
# GRADED FUNCTION: update_parameters
def update_parameters(parameters, grads, learning_rate = 1.2):
"""
Updates parameters using the gradient descent update rule given above
Arguments:
parameters -- python dictionary containing your parameters
grads -- python dictionary containing your gradients
Returns:
parameters -- python dictionary containing your updated parameters
"""
# Retrieve each parameter from the dictionary "parameters"
### START CODE HERE ### (≈ 4 lines of code)
W1 = parameters["W1"]
b1 = parameters["b1"]
W2 = parameters["W2"]
b2 = parameters["b2"]
### END CODE HERE ###
# Retrieve each gradient from the dictionary "grads"
### START CODE HERE ### (≈ 4 lines of code)
dW1 = grads["dW1"]
db1 = grads["db1"]
dW2 = grads["dW2"]
db2 = grads["db2"]
## END CODE HERE ###
# Update rule for each parameter
### START CODE HERE ### (≈ 4 lines of code)
W1 = W1 - learning_rate * dW1
b1 = b1 - learning_rate * db1
W2 = W2 - learning_rate * dW2
b2 = b2 - learning_rate * db2
### END CODE HERE ###
parameters = {"W1": W1,
"b1": b1,
"W2": W2,
"b2": b2}
return parameters
3.5 综合nn模型
# GRADED FUNCTION: nn_model
def nn_model(X, Y, n_h, num_iterations = 10000, print_cost=False):
"""
Arguments:
X -- dataset of shape (2, number of examples)
Y -- labels of shape (1, number of examples)
n_h -- size of the hidden layer
num_iterations -- Number of iterations in gradient descent loop
print_cost -- if True, print the cost every 1000 iterations
Returns:
parameters -- parameters learnt by the model. They can then be used to predict.
"""
np.random.seed(3)
n_x = layer_sizes(X, Y)[0]
n_y = layer_sizes(X, Y)[2]
# Initialize parameters, then retrieve W1, b1, W2, b2. Inputs: "n_x, n_h, n_y". Outputs = "W1, b1, W2, b2, parameters".
### START CODE HERE ### (≈ 5 lines of code)
parameters = initialize_parameters(n_x, n_h, n_y)
W1 = parameters["W1"]
b1 = parameters["b1"]
W2 = parameters["W2"]
b2 = parameters["b2"]
### END CODE HERE ###
# Loop (gradient descent)
for i in range(0, num_iterations):
### START CODE HERE ### (≈ 4 lines of code)
# Forward propagation. Inputs: "X, parameters". Outputs: "A2, cache".
A2, cache = forward_propagation(X, parameters)
# Cost function. Inputs: "A2, Y, parameters". Outputs: "cost".
cost = compute_cost(A2, Y, parameters)
# Backpropagation. Inputs: "parameters, cache, X, Y". Outputs: "grads".
grads = backward_propagation(parameters, cache, X, Y)
# Gradient descent parameter update. Inputs: "parameters, grads". Outputs: "parameters".
parameters = update_parameters(parameters, grads, learning_rate = 1.2)
### END CODE HERE ###
# Print the cost every 1000 iterations
if print_cost and i % 1000 == 0:
print ("Cost after iteration %i: %f" %(i, cost))
return parameters
3.6 预测
# GRADED FUNCTION: predict
def predict(parameters, X):
"""
Using the learned parameters, predicts a class for each example in X
Arguments:
parameters -- python dictionary containing your parameters
X -- input data of size (n_x, m)
Returns
predictions -- vector of predictions of our model (red: 0 / blue: 1)
"""
# Computes probabilities using forward propagation, and classifies to 0/1 using 0.5 as the threshold.
### START CODE HERE ### (≈ 2 lines of code)
A2, cache = forward_propagation(X, parameters)
predictions = (A2 > 0.5)
### END CODE HERE ###
return predictions
结果:
# Build a model with a n_h-dimensional hidden layer
parameters = nn_model(X, Y, n_h = 4, num_iterations = 10000, print_cost=True)
# Plot the decision boundary
plot_decision_boundary(lambda x: predict(parameters, x.T), X, Y)
plt.title("Decision Boundary for hidden layer size " + str(4))
小结:
1. 熟悉单层训练过程
2. 拓展:使用不同神经元个数的影响:
Accuracy for 1 hidden units: 67.5 %
Accuracy for 2 hidden units: 67.25 %
Accuracy for 3 hidden units: 90.75 %
Accuracy for 4 hidden units: 90.5 %
Accuracy for 5 hidden units: 91.25 %
Accuracy for 20 hidden units: 90.0 %
Accuracy for 50 hidden units: 90.25 %