浅谈pytorch卷积核大小的设置对全连接神经元的影响

脚本专栏 2024/11/25 佚名

3 2 1

3*3卷积核与2*5卷积核对神经元大小的设置

#这里kerner_size = 2*5
class CONV_NET(torch.nn.Module): #CONV_NET类继承nn.Module类
 def __init__(self):
  super(CONV_NET, self).__init__() #使CONV_NET类包含父类nn.Module的所有属性
  # super()需要两个实参，子类名和对象self
  self.conv1 = nn.Conv2d(1, 32, (2, 5), 1, padding=0)
  self.conv2 = nn.Conv2d(32, 128, 1, 1, padding=0)
  self.fc1 = nn.Linear(512, 128)
  self.relu1 = nn.ReLU(inplace=True)
  self.drop1 = nn.Dropout(0.5)
  self.fc2 = nn.Linear(128, 32)
  self.relu2 = nn.ReLU(inplace=True)
  self.fc3 = nn.Linear(32, 3)
  self.softmax = nn.Softmax(dim=1)

 def forward(self, x):
  x = self.conv1(x)
  x = self.conv2(x)
  x = x.view(x.size(0), -1)
  x = self.fc1(x)
  x = self.relu1(x)
  x = self.drop1(x)
  x = self.fc2(x)
  x = self.relu2(x)
  x = self.fc3(x)
  x = self.softmax(x)
  return x

主要看对称卷积核以及非对称卷积核之间的计算方式

#这里kerner_size = 3*3
class CONV_NET(torch.nn.Module): #CONV_NET类继承nn.Module类
 def __init__(self):
  super(CONV_NET, self).__init__() #使CONV_NET类包含父类nn.Module的所有属性
  # super()需要两个实参，子类名和对象self
  self.conv1 = nn.Conv2d(1, 32, 3, 1, padding=1)
  self.conv2 = nn.Conv2d(32, 128, 1, 1, padding=0)
  self.fc1 = nn.Linear(3200, 128)
  self.relu1 = nn.ReLU(inplace=True)
  self.drop1 = nn.Dropout(0.5)
  self.fc2 = nn.Linear(128, 32)
  self.relu2 = nn.ReLU(inplace=True)
  self.fc3 = nn.Linear(32, 3)
  self.softmax = nn.Softmax(dim=1)

 def forward(self, x):
  x = self.conv1(x)
  x = self.conv2(x)
  x = x.view(x.size(0), -1)
  x = self.fc1(x)
  x = self.relu1(x)
  x = self.drop1(x)
  x = self.fc2(x)
  x = self.relu2(x)
  x = self.fc3(x)
  x = self.softmax(x)
  return x

针对kerner_size=2*5,padding=0，stride=1以及kerner_size=3*3,padding=1,stride=1二者计算方式的比较如图所示