def gradient_check_n(parameters, gradients, X, Y, epsilon = 1e-7):
"""
Checks if backward_propagation_n computes correctly the gradient of the cost output by forward_propagation_n
Arguments:
parameters -- python dictionary containing your parameters "W1", "b1", "W2", "b2", "W3", "b3":
grad -- output of backward_propagation_n, contains gradients of the cost with respect to the parameters.
x -- input datapoint, of shape (input size, 1)
y -- true "label"
epsilon -- tiny shift to the input to compute approximated gradient with formula(1)
Returns:
difference -- difference (2) between the approximated gradient and the backward propagation gradient
"""
# Set-up variables
parameters_values, _ = dictionary_to_vector(parameters)
grad = gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))
gradapprox = np.zeros((num_parameters, 1))
# Compute gradapprox
for i in range(num_parameters):
# Compute J_plus[i]. Inputs: "parameters_values, epsilon". Output = "J_plus[i]".
# "_" is used because the function you have to outputs two parameters but we only care about the first one
### START CODE HERE ### (approx. 3 lines)
thetaplus = np.copy(parameters_values) # Step 1
thetaplus[i][0] = thetaplus[i] + epsilon # Step 2
J_plus[i], _ = forward_propagation_n(X, Y, vector_to_dictionary(thetaplus)) # Step 3
### END CODE HERE ###
# Compute J_minus[i]. Inputs: "parameters_values, epsilon". Output = "J_minus[i]".
### START CODE HERE ### (approx. 3 lines)
thetaminus = np.copy(parameters_values) # Step 1
thetaminus[i][0] = thetaminus[i] - epsilon # Step 2
J_minus[i], _ = forward_propagation_n(X, Y, vector_to_dictionary(thetaminus)) # Step 3
### END CODE HERE ###
# Compute gradapprox[i]
### START CODE HERE ### (approx. 1 line)
gradapprox[i] = (J_plus[i] - J_minus[i]) /(2 * epsilon)
### END CODE HERE ###
# Compare gradapprox to backward propagation gradients by computing difference.
### START CODE HERE ### (approx. 1 line)
numerator = np.linalg.norm(grad - gradapprox) # Step 1'
denominator = np.linalg.norm(grad) + np.linalg.norm(gradapprox) # Step 2'
difference = numerator / denominator # Step 3'
### END CODE HERE ###
if difference > 2e-7:
print ("\033[93m" + "There is a mistake in the backward propagation! difference = " + str(difference) + "\033[0m")
else:
print ("\033[92m" + "Your backward propagation works perfectly fine! difference = " + str(difference) + "\033[0m")
return difference
def gradient_check_n(parameters, gradients, X, Y, epsilon = 1e-7):
"""
Checks if backward_propagation_n computes correctly the gradient of the cost output by forward_propagation_n
Arguments:
parameters -- python dictionary containing your parameters "W1", "b1", "W2", "b2", "W3", "b3":
grad -- output of backward_propagation_n, contains gradients of the cost with respect to the parameters.
x -- input datapoint, of shape (input size, 1)
y -- true "label"
epsilon -- tiny shift to the input to compute approximated gradient with formula(1)
Returns:
difference -- difference (2) between the approximated gradient and the backward propagation gradient
"""
# Set-up variables
parameters_values, _ = dictionary_to_vector(parameters)
grad = gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))
gradapprox = np.zeros((num_parameters, 1))
e = epsilon
# Compute gradapprox
for i in range(num_parameters):
# Compute J_plus[i]. Inputs: "parameters_values, epsilon". Output = "J_plus[i]".
# "_" is used because the function you have to outputs two parameters but we only care about the first one
### START CODE HERE ### (approx. 3 lines)
thetaplus = np.copy(parameters_values) # Step 1
thetaplus[i][0] = thetaplus[i][0] + e # Step 2
J_plus[i], _ = forward_propagation_n(X, Y, vector_to_dictionary(thetaplus)) # Step 3
### END CODE HERE ###
# Compute J_minus[i]. Inputs: "parameters_values, epsilon". Output = "J_minus[i]".
### START CODE HERE ### (approx. 3 lines)
thetaminus = np.copy(parameters_values) # Step 1
thetaminus[i][0] = thetaminus[i][0] - e # Step 2
J_minus[i], _ = forward_propagation_n(X, Y, vector_to_dictionary(thetaminus)) # Step 3
### END CODE HERE ###
# Compute gradapprox[i]
### START CODE HERE ### (approx. 1 line)
gradapprox[i] = (J_plus[i] - J_minus[i]) / (2*e)
### END CODE HERE ###
# Compare gradapprox to backward propagation gradients by computing difference.
### START CODE HERE ### (approx. 1 line)
numerator = np.linalg.norm(gradapprox - grad)
denominator = np.linalg.norm(gradapprox) + np.linalg.norm(grad)
difference = numerator / denominator # Step 3'
### END CODE HERE ###
if difference > 2e-7:
print ("\033[93m" + "There is a mistake in the backward propagation! difference = " + str(difference) + "\033[0m")
else:
print ("\033[92m" + "Your backward propagation works perfectly fine! difference = " + str(difference) + "\033[0m")
return difference
def gradient_check_n(parameters,gradients,X,Y,epsilon=1e-7):
parameters_values, keys = gc_utils.dictionary_to_vector(parameters)
grad = gc_utils.gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))
gradapprox = np.zeros((num_parameters, 1))
for i in range(num_parameters):
# 计算J_plus[i]
theta_plus = np.copy(parameters_values)
theta_plus[i][0] = theta_plus[i][0] + epsilon
J_plus[i], cache = forward_propagation_n(X, Y, gc_utils.vector_to_dictionary(theta_plus))
# 计算J_minus[i]
theta_minus = np.copy(parameters_values)
theta_minus[i][0] = theta_minus[i][0] - epsilon
J_minus[i], cache = forward_propagation_n(X, Y, gc_utils.vector_to_dictionary(theta_minus))
# 计算gradapprox[i]
gradapprox[i] = (J_plus[i] - J_minus[i]) / (2 * epsilon)
numerator = np.linalg.norm(grad - gradapprox)
denominator = np.linalg.norm(grad) + np.linalg.norm(gradapprox)
difference = numerator / denominator
if difference < 1e-7:
print("梯度检查:梯度正常")
else:
print("梯度检测:梯度超出阈值")
return difference
def gradient_check_n(parameters, gradients, X, Y, epsilon=1e-7):
parameters_values, _ = dictionary_to_vector(parameters)
grad = gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))
gradapprox = np.zeros((num_parameters, 1))
for i in range(num_parameters):
thetaplus = np.copy(parameters_values)
thetaplus[i][0] = thetaplus[i][0] + epsilon
J_plus[i], _ = forward_propagation_n(X, Y,
vector_to_dictionary(thetaplus))
thetaminus = np.copy(parameters_values)
thetaminus[i][0] = thetaplus[i][0] - epsilon
J_minus[i], _ = forward_propagation_n(X, Y,
vector_to_dictionary(thetaminus))
gradapprox[i] = (J_plus[i] - J_minus[i]) / (2 * epsilon)
difference = np.linalg.norm(grad - gradapprox) / (
np.linalg.norm(grad) + np.linalg.norm(gradapprox))
if difference > 1e-7:
print("There is a mistake in the backward propagation! difference = " +
str(difference))
else:
print("Your backward propagation works perfectly fine! difference = " +
str(difference))
return difference
def gradient_check_n(parameters, gradients, X, Y, epsilon=1e-7):
parameters_values, _ = dictionary_to_vector(parameters)
grad = gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))
gradapprox = np.zeros((num_parameters, 1))
for i in range(num_parameters):
thetaplus = np.copy(parameters_values) # Step 1
thetaplus[i][0] = thetaplus[i][0] + epsilon # Step 2
J_plus[i], _ = forward_propagation_n(
X, Y, vector_to_dictionary(thetaplus)) # Step 3
thetaminus = np.copy(parameters_values) # Step 1
thetaminus[i][0] = thetaminus[i][0] - epsilon # Step 2
J_minus[i], _ = forward_propagation_n(
X, Y, vector_to_dictionary(thetaminus)) # Step 3
gradapprox[i] = (J_plus[i] - J_minus[i]) / (2 * epsilon)
numerator = np.linalg.norm(gradapprox - grad) # Step 1'
denominator = np.linalg.norm(grad) + np.linalg.norm(gradapprox) # Step 2'
difference = numerator / denominator # Step 3'
if difference > 1e-7:
print("\033[93m" +
"There is a mistake in the backward propagation! difference = " +
str(difference) + "\033[0m")
else:
print("\033[92m" +
"Your backward propagation works perfectly fine! difference = " +
str(difference) + "\033[0m")
return difference
def gradient_check_n(parameters, gradients, X, Y, epsilon=1e-7):
parameters_values, _ = dictionary_to_vector(parameters)
grad = gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))
gradapprox = np.zeros((num_parameters, 1))
for i in range(num_parameters):
thetaplus = np.copy(parameters_values)
thetaplus[i][0] = thetaplus[i][0] + epsilon
J_plus[i], _ = forward_propagation_n(X, Y,
vector_to_dictionary(thetaplus))
thetaminus = np.copy(parameters_values)
thetaminus[i][0] = thetaminus[i][0] - epsilon
J_minus[i], _ = forward_propagation_n(X, Y,
vector_to_dictionary(thetaminus))
gradapprox[i] = (J_plus[i] - J_minus[i]) / (2 * epsilon)
numerator = np.linalg.norm(grad - gradapprox)
denominator = np.linalg.norm(grad) + np.linalg.norm(gradapprox)
difference = numerator / denominator
if difference > 2e-7:
print("\033[93m" + "反向传播有问题! difference = " + str(difference) +
"\033[0m")
else:
print("\033[92m" + "反向传播很完美! difference = " + str(difference) +
"\033[0m")
return difference
def gradient_check_n(parameters, gradients, X, Y, epsilon=1e-7):
"""
Checks if backward_propagation_n computes correctly the gradient of the cost output by forward_propagation_n
Arguments:
parameters -- python dictionary containing your parameters "W1", "b1", "W2", "b2", "W3", "b3":
grad -- output of backward_propagation_n, contains gradients of the cost with respect to the parameters.
x -- input datapoint, of shape (input size, 1)
y -- true "label"
epsilon -- tiny shift to the input to compute approximated gradient.
Returns:
difference -- difference (2) between the approximated gradient and the backward propagation gradient.
"""
# Sets up variables
parameters_values, _ = dictionary_to_vector(parameters)
grad = gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))
gradapprox = np.zeros((num_parameters, 1))
# Computes gradapprox
for i in range(num_parameters):
# Computes J_plus[i].
thetaplus = np.copy(parameters_values)
thetaplus[i][0] = thetaplus[i][0] + epsilon
J_plus[i], _ = forward_propagation_n(X, Y,
vector_to_dictionary(thetaplus))
# Computes J_minus[i].
thetaminus = np.copy(parameters_values)
thetaminus[i][0] = thetaminus[i][0] - epsilon
J_minus[i], _ = forward_propagation_n(X, Y,
vector_to_dictionary(thetaminus))
# Computes gradapprox[i]
gradapprox[i] = (J_plus[i] - J_minus[i]) / (2 * epsilon)
# Compares gradapprox to backward propagation gradients by computing difference.
numerator = np.linalg.norm(grad - gradapprox)
denominator = np.linalg.norm(grad) + np.linalg.norm(gradapprox)
difference = numerator / denominator
if difference > 2e-7:
print("\033[93m" +
"There is a mistake in the backward propagation! difference = " +
str(difference) + "\033[0m")
else:
print("\033[92m" +
"Your backward propagation works perfectly fine! difference = " +
str(difference) + "\033[0m")
return difference
def gradient_check_n(parameters,gradients,X,Y,epsilon=1e-7):
"""
检查backward_propagation_n是否正确计算forward_propagation_n输出的成本梯度
参数：
parameters - 包含参数“W1”，“b1”,"W2","b2","W3","b3"的python字典
grentients - grad_output_propagation_n的输出 包含与参数相关的成本梯度
x - 输入数据点，维度为（输入节点数量，1）\
y - 标签
epsilon - 计算输入的微小偏移以计算近似梯度
返回：
difference - 近似梯度和后向传播梯度之间的差异
"""
#初始化参数
parameters_values,keys = gc_utils.dictionary_to_vector(parameters) #keys用不到
print("parameters"+str(parameters))
print("parameters_values"+str(parameters_values))
grad = gc_utils.gradients_to_vector(gradients)
print("gradients"+str(gradients))
print("grad"+str(grad))
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters,1))
J_minus = np.zeros((num_parameters,1))
gradapprox = np.zeros((num_parameters,1))
#计算gradapprox
for i in range(num_parameters):
#计算J——plus[i],输入："parameters_values,epsilon",输出"J_plus[i]"
thetaplus = np.copy(parameters_values) #strp1
thetaplus[i][0] = thetaplus[i][0] + epsilon #step2
J_plus[i],cache = forward_propagation_n(X,Y,gc_utils.vector_to_dictionary(thetaplus)) #step3 cache用不到
#计算J_minus[i].输入"parameters_values,epsilon",输出"J_minus[i]"
thetaminus = np.copy(parameters_values)
thetaminus[i][0] = thetaminus[i][0] - epsilon
J_minus[i],cache = forward_propagation_n(X,Y,gc_utils.vector_to_dictionary(thetaminus))
#计算gradapprox[i]
gradapprox[i] = (J_plus[i] - J_minus[i]) / (2 * epsilon)
#通过计算差异比较gradapprox和后向传播梯度
numerator = np.linalg.norm(grad - gradapprox)
denominator = np.linalg.norm(grad) + np.linalg.norm(gradapprox)
difference = numerator / denominator
if difference < 1e-7:
print("梯度检查：梯度正常！")
else:
print("梯度检查：梯度超出阈值！")
return difference
def gradient_check_n(parameters, gradients, X, Y, epsilon=1e-7):
"""
Checks if backward_propagation_n computes correctly the gradient of the cost output by forward_propagation_n
Arguments:
parameters -- python dictionary containing your parameters
grad -- output of backward_propagation_n, contains gradients of the cost with respect to the parameter s.
X -- input datapoint, of shape (input size, 1)
Y -- true "label"
epsilon -- tiny shift to the input to compute approximated gradient
Returns:
difference -- difference between approximated gradient and the backward propagation gradient
"""
# Set up variables
parameters_values, _ = dictionary_to_vector(parameters)
grad = gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))
gradapprox = np.zeros((num_parameters, 1))
# compute gradapprox
for i in range(num_parameters):
# Compute J_plus[i]. Inputs: "parameters_values, epsilon". Output: "J_plus[i]"
# "_" is used because the function you have to outputs two parameters but we only care about the first one
thetaplus = np.copy(parameters_values)
thetaplus[i][0] += epsilon
J_plus[i], _ = forward_propagation_n(X, Y,
vector_to_dictionary(thetaplus))
# Compute J_minus[i]. Inputs: "parameters_values, epsilon". Output: "J_minus[i]".
thetaminus = np.copy(parameters_values)
thetaminus[i][0] -= epsilon
J_minus[i], _ = forward_propagation_n(X, Y,
vector_to_dictionary(thetaminus))
# Compute gradapprox[i]
gradapprox[i] = (J_plus[i] - J_minus[i]) / (2 * epsilon)
# Compare gradapprox to backward propagation gradients by computing difference.
numerator = np.linalg.norm(gradapprox - grad)
denominator = np.linalg.norm(gradapprox) + np.linalg.norm(grad)
difference = numerator / denominator
if difference > 1.2e-7:
print("\033[93m" +
"There is a mistake in the backward propagation! difference = " +
str(difference) + "\033[0m")
else:
print("\033[92m" +
"Your backward propagation works perfectly fine! difference = " +
str(difference) + "\033[0m")
return difference
def gradient_check_n(parameters, gradients, X, Y, epsilon=1e-7):
"""
检查backward_propagation_n是否正确计算forward_propagation_n输出的成本梯度
参数：
parameters - 包含参数“W1”，“b1”，“W2”，“b2”，“W3”，“b3”的python字典：
grad_output_propagation_n的输出包含与参数相关的成本梯度。
x - 输入数据点，维度为（输入节点数量，1）
y - 标签
epsilon - 计算输入的微小偏移以计算近似梯度
返回：
difference - 近似梯度和后向传播梯度之间的差异
"""
# 初始化参数
parameters_values, keys = gc_utils.dictionary_to_vector(
parameters) # keys用不到
grad = gc_utils.gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))
gradapprox = np.zeros((num_parameters, 1))
# 计算gradapprox
for i in range(num_parameters):
# 计算J_plus [i]。输入：“parameters_values，epsilon”。输出=“J_plus [i]”
thetaplus = np.copy(parameters_values) # Step 1
thetaplus[i][0] = thetaplus[i][0] + epsilon # Step 2
J_plus[i], cache = forward_propagation_n(
X, Y, gc_utils.vector_to_dictionary(thetaplus)) # Step 3 ，cache用不到
# 计算J_minus [i]。输入：“parameters_values，epsilon”。输出=“J_minus [i]”。
thetaminus = np.copy(parameters_values) # Step 1
thetaminus[i][0] = thetaminus[i][0] - epsilon # Step 2
J_minus[i], cache = forward_propagation_n(
X, Y,
gc_utils.vector_to_dictionary(thetaminus)) # Step 3 ，cache用不到
# 计算gradapprox[i]
gradapprox[i] = (J_plus[i] - J_minus[i]) / (2 * epsilon)
# 通过计算差异比较gradapprox和后向传播梯度。
numerator = np.linalg.norm(grad - gradapprox) # Step 1'
denominator = np.linalg.norm(grad) + np.linalg.norm(gradapprox) # Step 2'
difference = numerator / denominator # Step 3'
if difference < 1e-7:
print("梯度检查：梯度正常!")
else:
print("梯度检查：梯度超出阈值!")
return difference
def gradient_check_n(parameters, gradients, X, Y, epsilon=1e-7):
"""
check if backward_propagation_n computes correctly the gradient of the cost output
arguments:
parameters -- dictionary containing your parameters "W1","b1","W2","b2","W3","b3"
grad -- output of backward_propagation_n
x -- input datapoint,shape(input size,1)
y -- true label
epsilon -- tiny shift to the input to compute approximated gradient
returns:
difference -- difference between the approximated gradient
"""
#set-up variables
parameters_values, _ = dictionary_to_vector(parameters)
grad = gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))
gradapprox = np.zeros((num_parameters, 1))
#compute gradapprox
for i in range(num_parameters):
thetaplus = np.copy(parameters_values)
thetaplus[i][0] = thetaplus[i][0] + epsilon
J_plus[i], _ = forward_propagation_n(X, Y,
vector_to_dictionary(thetaplus))
thetaminus = np.copy(parameters_values)
thetaminus[i][0] = thetaminus[i][0] - epsilon
J_minus[i], _ = forward_propagation_n(X, Y,
vector_to_dictionary(thetaminus))
gradapprox[i] = (J_plus[i] - J_minus[i]) / (2 * epsilon)
print('grad.shape = ', grad.shape)
print('gradapprox.shape = ', gradapprox.shape)
numerator = np.linalg.norm(grad - gradapprox)
denominator = np.linalg.norm(grad) + np.linalg.norm(gradapprox)
difference = numerator / denominator
if difference > 1e-7:
print('there is a mistake in the backward propagation, difference = ' +
str(difference))
else:
print('your backward propagation works perfectly fine! difference = ' +
str(difference))
return difference
def gradients_check(X, Y, lambd=0, keep_prob=1, init_method='he'):
layers_dims = [X.shape[0], 5, 3, 1]
# inintial params
if init_method == 'zeros':
params = init_zeros(layers_dims)
elif init_method == 'random':
params = init_random(layers_dims)
elif init_method == 'he':
params = init_he(layers_dims)
else:
print('Error: unexcepted init_method!')
# compute grads
a3, cache = forward_propagate_with_reg(X, params, keep_prob=keep_prob)
grads = backward_propagate_with_reg(X,
Y,
cache,
lambd=lambd,
keep_prob=keep_prob)
grads_vector = gc_utils.gradients_to_vector(grads)
theta, keys = gc_utils.dictionary_to_vector(params) #转化成向量方便索引(n, 1)
n = theta.shape[0] #参数个数
grads_approx_vector = np.zeros((n, 1))
# compute grads_approx
for i in range(n):
theta_p = np.copy(theta)
theta_p[i, 0] += 1e-7
params_p = gc_utils.vector_to_dictionary(theta_p)
theta_m = np.copy(theta)
theta_m[i, 0] -= 1e-7
params_m = gc_utils.vector_to_dictionary(theta_m)
a3_, cache_ = forward_propagate_with_reg(X,
params_p,
keep_prob=keep_prob)
J_p = compute_loss_with_reg(a3_, Y, params_p, lambd=lambd)
a3_, cache_ = forward_propagate_with_reg(X,
params_m,
keep_prob=keep_prob)
J_m = compute_loss_with_reg(a3_, Y, params_m, lambd=lambd)
d_approx = (J_p - J_m) / (2 * 1e-7)
grads_approx_vector[i, 0] = d_approx
# compute difference
numerator = np.linalg.norm(grads_vector - grads_approx_vector)
denominator = np.linalg.norm(grads_vector) + np.linalg.norm(
grads_approx_vector)
diff = numerator / denominator
return diff
def gradient_check_n(parameters, gradients, X, Y, epsilon=1e-7):
"""
Checks if backward_propagation_n computes correctly the gradient of the cost output by forward_propagation_n
Arguments:
parameters -- python dictionary containing your parameters "W1", "b1", "W2", "b2", "W3", "b3":
grad -- output of backward_propagation_n, contains gradients of the cost with respect to the parameters.
x -- input datapoint, of shape (input size, 1)
y -- true "label"
epsilon -- tiny shift to the input to compute approximated gradient with formula(1)
Returns:
difference -- difference (2) between the approximated gradient and the backward propagation gradient
"""
# Set-up variables
parameters_values, _ = dictionary_to_vector(parameters)
grad = gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))
gradapprox = np.zeros((num_parameters, 1))
# Compute gradapprox
for i in range(num_parameters):
# Compute J_plus[i]. Inputs: "parameters_values, epsilon". Output = "J_plus[i]".
# "_" is used because the function you have to outputs two parameters but we only care about the first one
### START CODE HERE ### (approx. 3 lines)
plus_copy = parameters_values.copy()
plus_copy[i] = plus_copy[i] + epsilon
J_plus[i], _ = forward_propagation_n(X, Y,
vector_to_dictionary(plus_copy))
# Compute J_minus[i]. Inputs: "parameters_values, epsilon". Output = "J_minus[i]".
minus_copy = parameters_values.copy()
minus_copy[i] = minus_copy[i] - epsilon
J_minus[i], _ = forward_propagation_n(X, Y,
vector_to_dictionary(minus_copy))
# Compute gradapprox[i]
### START CODE HERE ### (approx. 1 line)
gradapprox[i] = (J_plus[i] - J_minus[i]) / (2 * epsilon)
# Compare gradapprox to backward propagation gradients by computing difference.
num = np.linalg.norm(grad - gradapprox)
nom = np.linalg.norm(grad) + np.linalg.norm(gradapprox)
diff = num / nom
return diff
def gradient_check_n(parameters, gradients, X, Y, epsilon=1e-7):
"""
Checks if backward_propagation_n computes correctly the gradient of the cost output by forward_propagation_n
Arguments:
parameters -- python dictionary containing your parameters "W1", "b1", "W2", "b2", "W3", "b3":
grad -- output of backward_propagation_n, contains gradients of the cost with respect to the parameters.
x -- input datapoint, of shape (input size, 1)
y -- true "label"
epsilon -- tiny shift to the input to compute approximated gradient with formula(1)
Returns:
difference -- difference (2) between the approximated gradient and the backward propagation gradient
"""
parameters_values, _ = dictionary_to_vector(parameters)
grad = gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
Jplus = np.zeros((num_parameters, 1))
Jminus = np.zeros((num_parameters, 1))
gradapprox = np.zeros((num_parameters, 1))
for i in range(num_parameters):
thetaplus = np.copy(parameters_values)
thetaplus[i][0] += epsilon
Jplus[i], _ = forward_propagation_n(X, Y,
vector_to_dictionary(thetaplus))
thetaminus = np.copy(parameters_values)
thetaminus[i][0] -= epsilon
Jminus[i], _ = forward_propagation_n(X, Y,
vector_to_dictionary(thetaminus))
gradapprox[i] = (Jplus[i] - Jminus[i]) / (2. * epsilon)
diff = np.linalg.norm(grad - gradapprox) / (np.linalg.norm(grad) +
np.linalg.norm(gradapprox))
if diff > 1e-7:
print('There is a mistake in backword propagation diff = {}'.format(
diff))
else:
print(
'Your backward propagation works well with diff = {}'.format(diff))
return diff
def gradient_check_n(parameters, grads, X, Y, epsilon=1e-7):
'''
Checks if backward_propagation_n computes correctly the gradient of the cost output by forward_propagation_n
Arguments:
parameters -- python dictionary containing your parameters "W1", "b1", "W2", "b2", "W3", "b3":
grad -- output of backward_propagation_n, contains gradients of the cost with respect to the parameters.
x -- input datapoint, of shape (input size, 1)
y -- true "label"
epsilon -- tiny shift to the input to compute approximated gradient with formula(1)
Returns:
difference -- difference (2) between the approximated gradient and the backward propagation gradient
'''
parameters_values, _ = dictionary_to_vector(parameters)
grads_value = gradients_to_vector(grads)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))
gradsapprox = np.zeros((num_parameters, 1))
for i in range(num_parameters):
thetaplus = np.copy(parameters_values)
thetaplus[i][0] = thetaplus[i][0] + epsilon
J_plus[i], _ = forward_propagation_n(X, Y,
vector_to_dictionary(thetaplus))
thetaminus = np.copy(parameters_values)
thetaminus[i][0] = thetaminus[i][0] - epsilon
J_minus[i], _ = forward_propagation_n(X, Y,
vector_to_dictionary(thetaminus))
gradsapprox[i] = (J_plus[i] - J_minus[i]) / (2 * epsilon)
numeretor = np.linalg.norm(grads_value - gradsapprox)
denominator = np.linalg.norm(grads_value) + np.linalg.norm(gradsapprox)
difference = numeretor / denominator
if difference > 1e-7:
print(
'There is a mistake in the backward propagation! difference = {}'.
format(difference))
else:
print(
'Your backward propagation worlks perfectly fine! difference = {}'.
format(difference))
def gradient_check_n(parameters,gradients,X,Y,epsilon = 1e-7):
'''
检查backward_propagation_n是否正确计算forward_propagation_n输出的成本梯度
:param parameters: 包含参数'W1','b1','W2','b2','W3','b3'的python字典
grad_output_propagation_n的输出包含与参数相关的成本梯度
:param gradients:
:param X: 输入数据点,维度为(输入节点数量,1)
:param Y: 标签
:param epsilon计算输入的微小偏移以计算近似梯度:
:return:
近似梯度和后向传播梯度之间的差异
'''
#初始化参数
parameters_values , keys = gc_utils.dictionary_to_vector(parameters)#keys用不到,parameters_values是一个n行1列的矩阵
grad = gc_utils.gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters,1))
J_minus = np.zeros((num_parameters,1))
gradapprox = np.zeros(num_parameters,1)
#计算gradapprox
for i in range(num_parameters):
#计算J_plus[i].输入'parameters_values,epsilon'.输出='J_plus[i]'
thetaplus = np.copy(parameters_values)
thetaplus[i][0]=thetaplus[i][0]+epsilon
J_plus[i],cache = forward_propagation_n(X,Y,gc_utils.vector_to_dictionary(thetaplus))
#计算J_minus[i].输入:'parameters_values,epsilon',输出='J_minus[i]'
thetaminus = np.copy(parameters_values)
thetaminus[i][0]=thetaminus[i][0] - epsilon
J_minus[i],cache = forward_propagation_n(X,Y,gc_utils.vector_to_dictionary(thetaminus))
#计算gradapprox[i]
gradapprox[i] = (J_plus[i]-J_minus[i])/(2*epsilon)
#通过计算差异比较gradapprox和后向传播梯度
numerator = np.linalg.norm(grad-gradapprox)
denominator = np.linalg.norm(grad)+np.linalg.norm(gradapprox)
difference = numerator/denominator
if difference<1e-7:
print('梯度检查:梯度正常')
else:
print('梯度检查:梯度超出阈值')
return difference
def gradient_check_n(parameters, gradients, X, Y, epsilon = 1e-7):
# Set-up variables
parameters_values, _ = dictionary_to_vector(parameters)
grad = gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))
gradapprox = np.zeros((num_parameters, 1))
# Compute gradapprox
for i in range(num_parameters):
# Compute J_plus[i]. Inputs: "parameters_values, epsilon". Output = "J_plus[i]".
# "_" is used because the function you have to outputs two parameters but we only care about the first one
thetaplus = np.copy(parameters_values) # Step 1
thetaplus[i][0] += epsilon # Step 2
J_plus[i], _ = forward_propagation_n(X, Y, vector_to_dictionary(thetaplus)) # Step 3
# Compute J_minus[i]. Inputs: "parameters_values, epsilon". Output = "J_minus[i]".
thetaminus = np.copy(parameters_values) # Step 1
thetaminus[i][0] -= epsilon # Step 2
J_minus[i], _ = forward_propagation_n(X, Y, vector_to_dictionary(thetaminus)) # Step 3
# Compute gradapprox[i]
gradapprox[i] = (J_plus[i] - J_minus[i]) / (2. * epsilon)
# Compare gradapprox to backward propagation gradients by computing difference.
numerator = np.linalg.norm(grad - gradapprox) # Step 1'
denominator = np.linalg.norm(grad) + np.linalg.norm(gradapprox) # Step 2'
difference = numerator / denominator
if difference > 1e-6:
print ("There is a mistake in the backward propagation! difference = " + str(difference))
else:
print ("Your backward propagation works perfectly fine! difference = " + str(difference))
return difference
def gradient_check_n(parameters, gradients, X, Y, epsilon=1e-7):
# 初始化参数
parameters_values, keys = gc_utils.dictionary_to_vector(
parameters) # 将parameters字典转换为array
grad = gc_utils.gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))
gradapprox = np.zeros((num_parameters, 1))
# 计算grad approx
for i in range(num_parameters): # 遍历所有的参数
# 计算 J_plus[i]
thetaplus = np.copy(parameters_values)
thetaplus[i][0] = thetaplus[i][0] + epsilon
J_plus[i], cache = forward_propagation_n(
X, Y, gc_utils.vector_to_dictionary(thetaplus)) # cache用不到
# 计算 J_minus[i]
thetaminus = np.copy(parameters_values)
thetaminus[i][0] = thetaminus[i][0] - epsilon
J_minus[i], cache = forward_propagation_n(
X, Y, gc_utils.vector_to_dictionary(thetaminus)) # cache用不到
# 计算 grad apporx[i]
gradapprox[i] = (J_plus[i] - J_minus[i]) / (2 * epsilon)
# 通过计算差异比较 gradapprox 和后向传播梯度
numerator = np.linalg.norm(grad - gradapprox)
denominator = np.linalg.norm(grad) + np.linalg.norm(gradapprox)
difference = numerator / denominator
if difference < 1e-7:
print("Gradient Checking: 梯度正常！")
else:
print("Gradient Checking:梯度超出阈值！")
return difference
How does gradient checking work?. As in 1) and 2), you want to compare "gradapprox" to the gradient computed by backpropagation. The formula is still: ∂J∂θ=limε→0J(θ+ε)−J(θ−ε)2ε(1) (1)∂J∂θ=limε→0J(θ+ε)−J(θ−ε)2ε However, θθ is not a scalar anymore. It is a dictionary called "parameters". We implemented a function "dictionary_to_vector()" for you. It converts the "parameters" dictionary into a vector called "values", obtained by reshaping all parameters (W1, b1, W2, b2, W3, b3) into vectors and concatenating them. The inverse function is "vector_to_dictionary" which outputs back the "parameters" dictionary. Figure 2 : dictionary_to_vector() and vector_to_dictionary() You will need these functions in gradient_check_n() We have also converted the "gradients" dictionary into a vector "grad" using gradients_to_vector(). You don't need to worry about that. Exercise: Implement gradient_check_n(). Instructions: Here is pseudo-code that will help you implement the gradient check. For each i in num_parameters: To compute J_plus[i]: Set θ+θ+ to np.copy(parameters_values) Set θ+iθi+ to θ+i+εθi++ε Calculate J+iJi+ using to forward_propagation_n(x, y, vector_to_dictionary(θ+θ+ )). To compute J_minus[i]: do the same thing with θ−θ− Compute gradapprox[i]=J+i−J−i2εgradapprox[i]=Ji+−Ji−2ε Thus, you get a vector gradapprox, where gradapprox[i] is an approximation of the gradient with respect to parameter_values[i]. You can now compare this gradapprox vector to the gradients vector from backpropagation. Just like for the 1D case (Steps 1', 2', 3'), compute: difference=∥grad−gradapprox∥2∥grad∥2+∥gradapprox∥2(3)
def gradient_check_n(parameters, gradients, X, Y, epsilon=1e-7):
"""
Checks if backward_propagation_n computes correctly the gradient of the cost output by forward_propagation_n
Arguments:
parameters -- python dictionary containing your parameters "W1", "b1", "W2", "b2", "W3", "b3":
grad -- output of backward_propagation_n, contains gradients of the cost with respect to the parameters.
x -- input datapoint, of shape (input size, 1)
y -- true "label"
epsilon -- tiny shift to the input to compute approximated gradient with formula(1)
Returns:
difference -- difference (2) between the approximated gradient and the backward propagation gradient
"""
# Set-up variables
parameters_values, _ = dictionary_to_vector(parameters)
grad = gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))
gradapprox = np.zeros((num_parameters, 1))
# Compute gradapprox
for i in range(num_parameters):
# Compute J_plus[i]. Inputs: "parameters_values, epsilon". Output = "J_plus[i]".
# "_" is used because the function you have to outputs two parameters but we only care about the first one
thetaplus = np.copy(parameters_values) # Step 1
thetaplus[i][0] = thetaplus[i][0] + epsilon # Step 2
J_plus[i], _ = forward_propagation_n(
X, Y, vector_to_dictionary(thetaplus)) # Step 3
# Compute J_minus[i]. Inputs: "parameters_values, epsilon". Output = "J_minus[i]".
thetaminus = np.copy(parameters_values) # Step 1
thetaminus[i][0] = thetaminus[i][0] - epsilon # Step 2
J_minus[i], _ = forward_propagation_n(
X, Y, vector_to_dictionary(thetaminus)) # Step 3
# Compute gradapprox[i]
gradapprox[i] = (J_plus[i] - J_minus[i]) / (2 * epsilon)
# Compare gradapprox to backward propagation gradients by computing difference.
numerator = np.linalg.norm(grad - gradapprox) # Step 1'
denominator = np.linalg.norm(grad) + np.linalg.norm(gradapprox) # Step 2'
difference = numerator / denominator # Step 3'
if difference > 2e-7:
print("\033[93m" +
"There is a mistake in the backward propagation! difference = " +
str(difference) + "\033[0m")
else:
print("\033[92m" +
"Your backward propagation works perfectly fine! difference = " +
str(difference) + "\033[0m")
return difference
# What you should remember from this notebook:
# Gradient checking verifies closeness between the gradients from backpropagation
# and the numerical approximation of the gradient (computed using forward propagation).
# Gradient checking is slow, so we don't run it in every iteration of training.
# You would usually run it only to make sure your code is correct,
# then turn it off and use backprop for the actual learning process.
def gradient_check_n(parameters, gradients, X, Y, epsilon=1e-7):
"""
Checks if backward_propagation_n computes correctly the gradient of the cost output by forward_propagation_n
Arguments:
parameters -- python dictionary containing your parameters "W1", "b1", "W2", "b2", "W3", "b3":
grad -- output of backward_propagation_n, contains gradients of the cost with respect to the parameters.
x -- input datapoint, of shape (input size, 1)
y -- true "label"
epsilon -- tiny shift to the input to compute approximated gradient with formula(1)
Returns:
difference -- difference (2) between the approximated gradient and the backward propagation gradient
"""
# How to help implement gradient check.
# Set-up variables
parameters_values, _ = dictionary_to_vector(
parameters
) # converts the "parameters" dictionary into a vector called "values"
grad = gradients_to_vector(
gradients) # convert gradients dictionary into a vector, "grads"
num_parameters = parameters_values.shape[
0] # get current shape of an array by assigning a tuple of array dimensions
J_plus = np.zeros(
(num_parameters,
1)) # initialize J_plus with zeros and number of parameter objects
J_minus = np.zeros(
(num_parameters,
1)) # initialize J_minus with zeros and number of parameter objects
gradapprox = np.zeros((
num_parameters,
1)) # initialize gradapprox with zeros and number of parameter objects
# Compute gradapprox
for i in range(num_parameters):
# Compute J_plus[i]. Inputs: "parameters_values, epsilon". Output = "J_plus[i]".
# "_" is used because the function outputs two parameters but only care about the first one
thetaplus = np.copy(parameters_values) # Set theta to np.copy
thetaplus[i][0] = thetaplus[i][0] + epsilon # Set theta_plus
J_plus[i], _ = forward_propagation_n(
X, Y, vector_to_dictionary(
thetaplus)) # Calculate J_plus using forward propagation
# Compute J_minus[i]. Inputs: "parameters_values, epsilon". Output = "J_minus[i]".
thetaminus = np.copy(parameters_values) # Set theta to np.copy
thetaminus[i][0] = thetaminus[i][0] - epsilon # Set theta_minus
J_minus[i], _ = forward_propagation_n(
X, Y, vector_to_dictionary(
thetaminus)) # Calculate J_minus using forward propagation
# Compute gradapprox[i]
gradapprox[i] = (J_plus[i] - J_minus[i]) / (2 * epsilon)
# Compare gradapprox to backward propagation gradients by computing difference.
numerator = np.linalg.norm(
grad - gradapprox) # compute the numerator using np.linag.norm(...)
denominator = np.linalg.norm(grad) + np.linalg.norm(
gradapprox
) # compute the denominator(need to call np.linag.norm(...) twice)
difference = numerator / denominator # divide both
if difference > 2e-7:
print("\033[93m" +
"There is a mistake in the backward propagation! difference = " +
str(difference) + "\033[0m")
else:
print("\033[92m" +
"Your backward propagation works perfectly fine! difference = " +
str(difference) + "\033[0m")
return difference
def gradient_check_n(parameters, gradients, X, Y, epsilon=1e-7):
"""
梯度校验，检查后向传播是否正确计算前向传播输出的 cost 的梯度
:param parameters: 参数字典，包含 "W1", "b1", "W2", "b2", "W3", "b3":
:param grad: 后向传播的输出, 包含与参数相关的 cost 梯度
:param x: 输入数据点, of shape (input size, 1)
:param y: 正确的标签
:param epsilon: 输入的微小偏移，用来计算近似梯度
:return difference: 近似梯度和后向传播计算的梯度之间的差异
"""
# Set-up variables
parameters_values, _ = dictionary_to_vector(parameters)
grad = gradients_to_vector(gradients)
num_parameters = parameters_values.shape[0]
J_plus = np.zeros((num_parameters, 1))
J_minus = np.zeros((num_parameters, 1))
gradapprox = np.zeros((num_parameters, 1))
# 计算 gradapprox
for i in range(num_parameters):
# Compute J_plus[i]. Inputs: "parameters_values, epsilon". Output = "J_plus[i]".
# "_" is used because the function you have to outputs two parameters but we only care about the first one
### START CODE HERE ### (approx. 3 lines)
thetaplus = np.copy(parameters_values) # Step 1
thetaplus[i][0] = thetaplus[i][0] + epsilon # Step 2
J_plus[i], _ = forward_propagation_n(
X, Y, vector_to_dictionary(thetaplus)) # Step 3
### END CODE HERE ###
# Compute J_minus[i]. Inputs: "parameters_values, epsilon". Output = "J_minus[i]".
### START CODE HERE ### (approx. 3 lines)
thetaminus = np.copy(parameters_values) # Step 1
thetaminus[i][0] = thetaminus[i][0] - epsilon # Step 2
J_minus[i], _ = forward_propagation_n(
X, Y, vector_to_dictionary(thetaminus)) # Step 3
### END CODE HERE ###
# Compute gradapprox[i]
### START CODE HERE ### (approx. 1 line)
gradapprox[i] = (J_plus[i] - J_minus[i]) / (2 * epsilon)
### END CODE HERE ###
# 通过计算差异来比较 gradapprox 梯度和后向传播计算的梯度
### START CODE HERE ### (approx. 1 line)
numerator = np.linalg.norm(grad - gradapprox) # Step 1'
denominator = np.linalg.norm(grad) + np.linalg.norm(gradapprox) # Step 2'
difference = numerator / denominator # Step 3'
### END CODE HERE ###
if difference > 2e-7:
print("\033[93m" +
"There is a mistake in the backward propagation! difference = " +
str(difference) + "\033[0m")
else:
print("\033[92m" +
"Your backward propagation works perfectly fine! difference = " +
str(difference) + "\033[0m")
return difference
