- 构建CNN中padding(填充)(只会对长和宽进行填充)
# GRADED FUNCTION: zero_pad
def zero_pad(X, pad):
"""
Pad with zeros all images of the dataset X. The padding is applied to the height and width of an image,
as illustrated in Figure 1.
Argument:
X -- python numpy array of shape (m, n_H, n_W, n_C) representing a batch of m images
pad -- integer, amount of padding around each image on vertical and horizontal dimensions
Returns:
X_pad -- padded image of shape (m, n_H + 2 * pad, n_W + 2 * pad, n_C)
"""
#(≈ 1 line)
# X_pad = None
# YOUR CODE STARTS HERE
# print(X.shape)
X_pad=np.pad(X,((0,0),(pad,pad),(pad,pad),(0,0)),mode='constant',constant_values=(0,0))
# YOUR CODE ENDS HERE
return X_pad
- 卷积运算(事实上,并不是数学中的卷积运算,而是逐个相乘,最后相加,element-wise 运算)
# GRADED FUNCTION: conv_single_step
def conv_single_step(a_slice_prev, W, b):
"""
Apply one filter defined by parameters W on a single slice (a_slice_prev) of the output activation
of the previous layer.
Arguments:
a_slice_prev -- slice of input data of shape (f, f, n_C_prev)
W -- Weight parameters contained in a window - matrix of shape (f, f, n_C_prev)
b -- Bias parameters contained in a window - matrix of shape (1, 1, 1)
Returns:
Z -- a scalar value, the result of convolving the sliding window (W, b) on a slice x of the input data
"""
#(≈ 3 lines of code)
# Element-wise product between a_slice_prev and W. Do not add the bias yet.
# s = None
# Sum over all entries of the volume s.
# Z = None
# Add bias b to Z. Cast b to a float() so that Z results in a scalar value.
# Z = None
# YOUR CODE STARTS HERE
s = a_slice_prev*W
Z = np.sum(s)
Z = Z + np.squeeze(b)
# YOUR CODE ENDS HERE
return Z
- CN层的实现(并不包含激活函数)
# GRADED FUNCTION: conv_forward
def conv_forward(A_prev, W, b, hparameters):
"""
Implements the forward propagation for a convolution function
Arguments:
A_prev -- output activations of the previous layer,
numpy array of shape (m, n_H_prev, n_W_prev, n_C_prev)
W -- Weights, numpy array of shape (f, f, n_C_prev, n_C)
b -- Biases, numpy array of shape (1, 1, 1, n_C)
hparameters -- python dictionary containing "stride" and "pad"
Returns:
Z -- conv output, numpy array of shape (m, n_H, n_W, n_C)
cache -- cache of values needed for the conv_backward() function
"""
# Retrieve dimensions from A_prev's shape (≈1 line)
# (m, n_H_prev, n_W_prev, n_C_prev) = None
(m, n_H_prev, n_W_prev, n_C_prev) = A_prev.shape
# Retrieve dimensions from W's shape (≈1 line)
# (f, f, n_C_prev, n_C) = None
(f, f, n_C_prev, n_C) = W.shape
# Retrieve information from "hparameters" (≈2 lines)
# stride = None
# pad = None
stride = hparameters['stride']
pad = hparameters['pad']
# Compute the dimensions of the CONV output volume using the formula given above.
# Hint: use int() to apply the 'floor' operation. (≈2 lines)
# n_H = None
# n_W = None
n_H = int((n_H_prev-f+2*pad)/stride)+1
n_W = int((n_W_prev-f+2*pad)/stride)+1
# Initialize the output volume Z with zeros. (≈1 line)
# Z = None
Z = np.zeros((m,n_H,n_W,n_C))
# Create A_prev_pad by padding A_prev
# A_prev_pad = None
A_prev_pad = zero_pad(A_prev,pad)
# for i in range(None): # loop over the batch of training examples
# a_prev_pad = None # Select ith training example's padded activation
# for h in range(None): # loop over vertical axis of the output volume
# Find the vertical start and end of the current "slice" (≈2 lines)
# vert_start = None
# vert_end = None
# for w in range(None): # loop over horizontal axis of the output volume
# Find the horizontal start and end of the current "slice" (≈2 lines)
# horiz_start = None
# horiz_end = None
# for c in range(None): # loop over channels (= #filters) of the output volume
# Use the corners to define the (3D) slice of a_prev_pad (See Hint above the cell). (≈1 line)
# a_slice_prev = None
# Convolve the (3D) slice with the correct filter W and bias b, to get back one output neuron. (≈3 line)
# weights = None
# biases = None
# Z[i, h, w, c] = None
# YOUR CODE STARTS HERE
for i in range(m):
a_prev_pad = A_prev_pad[i]
for h in range(n_H):
vert_start = stride * h
vert_end = stride * h + f
for w in range(n_W):
horiz_start = stride*w
horiz_end = stride*w+f
for c in range(n_C):
a_slice_prev = a_prev_pad[vert_start:vert_end,horiz_start:horiz_end,:]
weight = W[:,:,:,c]
#print('a'a_slice_prev)
#print(W.shape)
#print(weight.shape)
biases = b[:,:,:,c]
Z[i,h,w,c] = conv_single_step(a_slice_prev,weight,biases)
# YOUR CODE ENDS HERE
# print(Z[0,:,:,0])
# print("A_prev",A_prev[0])
# print("A",A_prev_pad[0])
# print(W_prev[:,:,:,0])
# print(b[:,:,:,0])
# Save information in "cache" for the backprop
cache = (A_prev, W, b, hparameters)
return Z, cache
- Pool层的实现
# GRADED FUNCTION: pool_forward
def pool_forward(A_prev, hparameters, mode = "max"):
"""
Implements the forward pass of the pooling layer
Arguments:
A_prev -- Input data, numpy array of shape (m, n_H_prev, n_W_prev, n_C_prev)
hparameters -- python dictionary containing "f" and "stride"
mode -- the pooling mode you would like to use, defined as a string ("max" or "average")
Returns:
A -- output of the pool layer, a numpy array of shape (m, n_H, n_W, n_C)
cache -- cache used in the backward pass of the pooling layer, contains the input and hparameters
"""
# Retrieve dimensions from the input shape
(m, n_H_prev, n_W_prev, n_C_prev) = A_prev.shape
# Retrieve hyperparameters from "hparameters"
f = hparameters["f"]
stride = hparameters["stride"]
# Define the dimensions of the output
n_H = int(1 + (n_H_prev - f) / stride)
n_W = int(1 + (n_W_prev - f) / stride)
n_C = n_C_prev
# Initialize output matrix A
A = np.zeros((m, n_H, n_W, n_C))
# for i in range(None): # loop over the training examples
# for h in range(None): # loop on the vertical axis of the output volume
# Find the vertical start and end of the current "slice" (≈2 lines)
# vert_start = None
# vert_end = None
# for w in range(None): # loop on the horizontal axis of the output volume
# Find the vertical start and end of the current "slice" (≈2 lines)
# horiz_start = None
# horiz_end = None
# for c in range (None): # loop over the channels of the output volume
# Use the corners to define the current slice on the ith training example of A_prev, channel c. (≈1 line)
# a_prev_slice = None
# Compute the pooling operation on the slice.
# Use an if statement to differentiate the modes.
# Use np.max and np.mean.
# if mode == "max":
# A[i, h, w, c] = None
# elif mode == "average":
# A[i, h, w, c] = None
# YOUR CODE STARTS HERE
for i in range(m):
for h in range(n_H):
vert_start = h*stride
vert_end = h*stride+f
for w in range(n_W):
horiz_start = w*stride
horiz_end = horiz_start+f
for c in range(n_C):
a_prev_slice = A_prev[i,vert_start:vert_end,horiz_start:horiz_end,c]
if mode=='max':
A[i,h,w,c] = np.max(a_prev_slice)
elif mode == 'average':
A[i,h,w,c] = np.mean(a_prev_slice)
# YOUR CODE ENDS HERE
# Store the input and hparameters in "cache" for pool_backward()
cache = (A_prev, hparameters)
# Making sure your output shape is correct
#assert(A.shape == (m, n_H, n_W, n_C))
return A, cache
下面进行CNN的反向传播
对于绝大多数的深度学习框架,只需要进行前向传播,便可以自动实现反向传播,但是我们依然可以手动实现反向传播。
- CN层的反向传播,这些求导知识需要用到矩阵的求导,可以参考《The Matrix coodbook》这本书
d A + = ∑ h = 0 n H ∑ w = 0 n W W c × d Z h w (1) dA \mathrel{+}= \sum _{h=0} ^{n_H} \sum_{w=0} ^{n_W} W_c \times dZ_{hw} \tag{1} dA+=h=0∑nHw=0∑nWWc×dZhw(1)
d W c + = ∑ h = 0 n H ∑ w = 0 n W a s l i c e × d Z h w (2) dW_c \mathrel{+}= \sum _{h=0} ^{n_H} \sum_{w=0} ^ {n_W} a_{slice} \times dZ_{hw} \tag{2} dWc+=h=0∑nHw=0∑nWaslice×dZhw(2)
d b = ∑ h ∑ w d Z h w (3) db = \sum_h \sum_w dZ_{hw} \tag{3} db=h∑w∑dZhw(3)
def conv_backward(dZ, cache):
"""
Implement the backward propagation for a convolution function
Arguments:
dZ -- gradient of the cost with respect to the output of the conv layer (Z), numpy array of shape (m, n_H, n_W, n_C)
cache -- cache of values needed for the conv_backward(), output of conv_forward()
Returns:
dA_prev -- gradient of the cost with respect to the input of the conv layer (A_prev),
numpy array of shape (m, n_H_prev, n_W_prev, n_C_prev)
dW -- gradient of the cost with respect to the weights of the conv layer (W)
numpy array of shape (f, f, n_C_prev, n_C)
db -- gradient of the cost with respect to the biases of the conv layer (b)
numpy array of shape (1, 1, 1, n_C)
"""
# Retrieve information from "cache"
# (A_prev, W, b, hparameters) = None
# Retrieve dimensions from A_prev's shape
# (m, n_H_prev, n_W_prev, n_C_prev) = None
# Retrieve dimensions from W's shape
# (f, f, n_C_prev, n_C) = None
(A_prev, W, b, hparameters) = cache
(m, n_H_prev, n_W_prev, n_C_prev) = A_prev.shape
(f, f, n_C_prev, n_C) = W.shape
# Retrieve information from "hparameters"
# stride = None
# pad = None
stride = hparameters['stride']
pad = hparameters['pad']
# Retrieve dimensions from dZ's shape
# (m, n_H, n_W, n_C) = None
(m, n_H, n_W, n_C) = dZ.shape
# Initialize dA_prev, dW, db with the correct shapes
# dA_prev = None
# dW = None
# db = None
dA_prev = np.zeros(shape=A_prev.shape)
dW = np.zeros(shape=W.shape)
db = np.zeros((1,1,1,n_C))
# Pad A_prev and dA_prev
# A_prev_pad = zero_pad(A_prev, pad)
# dA_prev_pad = zero_pad(dA_prev, pad)
A_prev_pad = zero_pad(A_prev,pad)
dA_prev_pad = zero_pad(dA_prev,pad)
#for i in range(m): # loop over the training examples
# select ith training example from A_prev_pad and dA_prev_pad
# a_prev_pad = None
# da_prev_pad = None
#for h in range(n_H): # loop over vertical axis of the output volume
# for w in range(n_W): # loop over horizontal axis of the output volume
# for c in range(n_C): # loop over the channels of the output volume
# Find the corners of the current "slice"
# vert_start = None
# vert_end = None
# horiz_start = None
# horiz_end = None
# Use the corners to define the slice from a_prev_pad
# a_slice = None
# Update gradients for the window and the filter's parameters using the code formulas given above
# da_prev_pad[vert_start:vert_end, horiz_start:horiz_end, :] += None
# dW[:,:,:,c] += None
# db[:,:,:,c] += None
# Set the ith training example's dA_prev to the unpadded da_prev_pad (Hint: use X[pad:-pad, pad:-pad, :])
# dA_prev[i, :, :, :] = None
# YOUR CODE STARTS HERE
for i in range(m):
a_prev_pad = A_prev_pad[i,:,:,:]
da_prev_pad = dA_prev_pad[i,:,:,:]
for h in range(n_H):
for w in range(n_W):
for c in range(n_C):
vert_start = stride * h
vert_end = vert_start + f
horiz_start = stride * w
horiz_end = horiz_start + f
a_slice = a_prev_pad[vert_start:vert_end,horiz_start:horiz_end,:]
# print(a_slice.shape)
da_prev_pad[vert_start:vert_end,horiz_start:horiz_end,:] += W[:,:,:,c]*dZ[i,h,w,c]
dW[:,:,:,c]+=a_slice*dZ[i,h,w,c]
db[:,:,:,c]+=dZ[i,h,w,c]
dA_prev[i,:,:,:] = da_prev_pad[pad:-pad,pad:-pad,:]
# YOUR CODE ENDS HERE
# Making sure your output shape is correct
assert(dA_prev.shape == (m, n_H_prev, n_W_prev, n_C_prev))
return dA_prev, dW, db
- 尽管pool层没有任何参数,但是我们依然需要考虑他的影响,并且对于Max和average还要进行不同的处理
MaxPooling将该梯度应用于值最高的坐标中
def create_mask_from_window(x):
"""
Creates a mask from an input matrix x, to identify the max entry of x.
Arguments:
x -- Array of shape (f, f)
Returns:
mask -- Array of the same shape as window, contains a True at the position corresponding to the max entry of x.
"""
# (≈1 line)
# mask = None
# YOUR CODE STARTS HERE
biggest = np.max(x)
mask = (x==biggest)
# YOUR CODE ENDS HERE
return mask
average是将该梯度平均应用于(f*f)矩阵中
def distribute_value(dz, shape):
"""
Distributes the input value in the matrix of dimension shape
Arguments:
dz -- input scalar
shape -- the shape (n_H, n_W) of the output matrix for which we want to distribute the value of dz
Returns:
a -- Array of size (n_H, n_W) for which we distributed the value of dz
"""
# Retrieve dimensions from shape (≈1 line)
# (n_H, n_W) = None
(n_H, n_W) = shape
# Compute the value to distribute on the matrix (≈1 line)
# average = None
average = dz/(n_H*n_W)
# Create a matrix where every entry is the "average" value (≈1 line)
# a = None
# YOUR CODE STARTS HERE
a = np.ones((n_H,n_W))*average
# YOUR CODE ENDS HERE
return a
def pool_backward(dA, cache, mode = "max"):
"""
Implements the backward pass of the pooling layer
Arguments:
dA -- gradient of cost with respect to the output of the pooling layer, same shape as A
cache -- cache output from the forward pass of the pooling layer, contains the layer's input and hparameters
mode -- the pooling mode you would like to use, defined as a string ("max" or "average")
Returns:
dA_prev -- gradient of cost with respect to the input of the pooling layer, same shape as A_prev
"""
# Retrieve information from cache (≈1 line)
# (A_prev, hparameters) = None
# Retrieve hyperparameters from "hparameters" (≈2 lines)
# stride = None
# f = None
# Retrieve dimensions from A_prev's shape and dA's shape (≈2 lines)
# m, n_H_prev, n_W_prev, n_C_prev = None
# m, n_H, n_W, n_C = None
# Initialize dA_prev with zeros (≈1 line)
# dA_prev = None
# for i in range(None): # loop over the training examples
# select training example from A_prev (≈1 line)
# a_prev = None
# for h in range(n_H): # loop on the vertical axis
# for w in range(n_W): # loop on the horizontal axis
# for c in range(n_C): # loop over the channels (depth)
# Find the corners of the current "slice" (≈4 lines)
# vert_start = None
# vert_end = None
# horiz_start = None
# horiz_end = None
# Compute the backward propagation in both modes.
# if mode == "max":
# Use the corners and "c" to define the current slice from a_prev (≈1 line)
# a_prev_slice = None
# Create the mask from a_prev_slice (≈1 line)
# mask = None
# Set dA_prev to be dA_prev + (the mask multiplied by the correct entry of dA) (≈1 line)
# dA_prev[i, vert_start: vert_end, horiz_start: horiz_end, c] += None
# elif mode == "average":
# Get the value da from dA (≈1 line)
# da = None
# Define the shape of the filter as fxf (≈1 line)
# shape = None
# Distribute it to get the correct slice of dA_prev. i.e. Add the distributed value of da. (≈1 line)
# dA_prev[i, vert_start: vert_end, horiz_start: horiz_end, c] += None
# YOUR CODE STARTS HERE
(A_prev, hparameters) = cache
stride = hparameters['stride']
f = hparameters['f']
m, n_H_prev, n_W_prev, n_C_prev = A_prev.shape
m, n_H, n_W, n_C = dA.shape
dA_prev = np.zeros(A_prev.shape)
for i in range(m):
a_prev = A_prev[i,:,:,:]
for h in range(n_H):
for w in range(n_W):
for c in range(n_C):
vert_start = h*stride
vert_end = vert_start+f
horiz_start = w*stride
horiz_end = horiz_start+f
if mode == 'max':
a_prev_slice = a_prev[vert_start:vert_end,horiz_start:horiz_end,c]
mask = create_mask_from_window(a_prev_slice)
# print(mask)
dA_prev[i,vert_start:vert_end,horiz_start:horiz_end,c] += mask*dA[i,h,w,c]
elif mode == 'average':
da = dA[i,h,w,c]
shape=(f,f)
dA_prev[i,vert_start:vert_end,horiz_start:horiz_end,c] +=distribute_value(da,shape)
# YOUR CODE ENDS HERE
# Making sure your output shape is correct
assert(dA_prev.shape == A_prev.shape)
return dA_prev