3.1 common neural network layer (I) image correlation layer

Image correlation layer mainly includes convolution layer （Conv)、 Pooling layer （Pool） etc. , These layers can be divided into one dimension in practical use （1Ｄ）、 A two-dimensional （２Ｄ） And 3D （３Ｄ）, Pooling mode is divided into average pooling （ＡvgPool)、 Maximum pooling （ＭaxPool)、 Adaptive pooling (AdaptiveAvgPool) etc. . Convolution layer in addition to the commonly used forward convolution , And then there's deconvolution （ＴransposeConv).

from PIL import Image
from torchvision.transforms import ToTensor, ToPILImage
import matplotlib.pyplot as plt
import torch as t
import torch.nn as nn
import torch.nn.functional as F
from torch.autograd import Variable as V
to_tensor=ToTensor()#image-->tensor
to_pil=ToPILImage()# and transforms.ToPILImage Will be Tensor or numpy.ndarray Turn into PIL.Image
lena=Image.open('/home/yangyuting/Desktop/20180925183958930.jpeg')
lena = lena.convert('L')# Convert the image to grayscale 
plt.figure("pic")
plt.imshow(lena)
plt.show()

The image after convolution of the image ：

# The input is a batch,batch_size=1
input=to_tensor(lena).unsqueeze(0)
# Sharpen the convolution kernel 
kernel=t.ones(3,3)/-9
kernel[1][1]=1＃ The convolution kernel is centered on １, Surrounded by １／９ Convolution kernel 
conv=nn.Conv2d(1,1,(3,3),1,bias=False)＃（in_channels, out_channels, kernel_size, stride=1）
conv.weight.data=kernel.view(1,1,3,3)#(batch_size,in_channels, kernel_size)
out=conv(V(input))
to_pil(out.data.squeeze(0))
plt.figure("pic")
plt.imshow(to_pil(out.data.squeeze(0)))
plt.show()

Linear： Fully connected layer

# Input batch_size=2, dimension =3
input=V(t.randn(2,3))
linear=nn.Linear(3,4)
h=linear(input)
print(h)

ＢatchNorm: Batch normalization layer , It is divided into １D,２D, 3 D

#4channel, The initialization standard deviation is 4, The mean for 0
bn=nn.BatchNorm1d(4)
bn.weight.data=t.ones(4)*4
bn.bias.data=t.zeros(4)

bn_out=bn(h)
bn_out.mean(0),bn_out.var(0,unbiased=False)

Dropout:dropout layer , To prevent over fitting , Also divided into １D,２D, 3 D

from PIL import Image
from torchvision.transforms import ToTensor, ToPILImage
import matplotlib.pyplot as plt
import torch as t
import torch.nn as nn
import torch.nn.functional as F
from torch.autograd import Variable as V

to_tensor=ToTensor()#image-->tensor
to_pil=ToPILImage()
lena=Image.open('/home/yangyuting/Desktop/20180925183958930.jpeg')
lena = lena.convert('L')
# plt.figure("pic")
# plt.imshow(lena)
# plt.show()

# The input is a batch,batch_size=1
input=to_tensor(lena).unsqueeze(0)
# Sharpen the convolution kernel 
kernel=t.ones(3,3)/-9
kernel[1][1]=1
conv=nn.Conv2d(1,1,(3,3),1,bias=False)
conv.weight.data=kernel.view(1,1,3,3)
out=conv(V(input))
to_pil(out.data.squeeze(0))
# plt.figure("pic")
# plt.imshow(to_pil(out.data.squeeze(0)))
# plt.show()

pool=nn.AvgPool2d(2,2)
list(pool.parameters())

out=pool(V(input))
to_pil(out.data.squeeze(0))
# plt.figure("pic")
# plt.imshow(to_pil(out.data.squeeze(0)))
# plt.show()
# Fully connected layer Linear
# Input batch_size=2, dimension =3
input=V(t.randn(2,3))
linear=nn.Linear(3,4)
h=linear(input)
#print(h)
#4channel, The initialization standard deviation is 4, The mean for 0
bn=nn.BatchNorm1d(4)
bn.weight.data=t.ones(4)*4
bn.bias.data=t.zeros(4)

bn_out=bn(h)
bn_out.mean(0),bn_out.var(0,unbiased=False)
print(bn_out)
# Each element is represented by 0.5 The probability of abandoning ＃ About half of the numbers become ０
dropout=nn.Dropout(0.5)
O=dropout(bn_out)
print(O)

m = nn.Dropout (p=0.2)
input = t.randn (3, 4)
print(input)
output = m (input)
print(output)

Output ：

tensor([[-3.9998, -3.9999,  3.9996,  3.9998],
        [ 3.9998,  3.9999, -3.9996, -3.9998]],
       grad_fn=<NativeBatchNormBackward>)
tensor([[-0.0000, -7.9997,  7.9993,  0.0000],
        [ 0.0000,  7.9997, -0.0000, -0.0000]], grad_fn=<MulBackward0>)
tensor([[-0.6623,  1.3807,  2.1652, -0.1931],
        [ 0.6697, -0.2446,  1.3638,  0.5675],
        [ 1.9312,  0.8611,  0.7551,  0.2626]])
tensor([[-0.8278,  1.7259,  2.7066, -0.2414],
        [ 0.8371, -0.0000,  1.7047,  0.7094],
        [ 2.4140,  0.0000,  0.9439,  0.0000]])

Be careful ：input Part of it becomes 0 in the future , Other vectors are scaled , That is to multiply by 1/(1-p). Therefore, it will be found that during the experiment ,input and output It's not a simple partial zero . How to explain in the image remains to be studied ？