学习笔记2：Pytorch 神经网络初始化权重与偏置

class CONCAT_CNN(nn.Module):
    def __init__(self):
        super(CONCAT_CNN, self).__init__()
        self.conv1_1 = nn.Conv2d(1, 64, kernel_size=3, stride=1, padding=1)
    
    def forward(self, x):
        conv1_1 = F.relu(self.conv1_1(x))
        return conv1_1

    def initialize(self):
        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                # nn.init.normal_(m.weight.data)  # normal: mean=0, std=1
                nn.init.kaiming_normal_(m.weight.data)
                m.bias.data.fill_(0)
net=CONCAT_CNN()
net.initialize()
print(net)
print("Total number of paramerters in networks is {}  ".format(sum(x.numel() for x in net.parameters())))

正态（normal）均匀( uniform)
(适用于RELU)kaiming正态、kaiming均匀

nn.init.kaiming_normal_(m.weight.data)
nn.init.kaiming_uniform_(m.weight.data)

均匀分布、正态分布（高斯分布）

nn.init.uniform_(m.weight.data)# normal: mean=0, std=1
nn.init.normal_(m.weight.data,std=0.01)

（适用于Sigmoid，Tanh）Xavier正态、均匀

nn.init.xavier_normal_(m.weight.data)
nn.init.xavier_uniform_(m.weight.data)