批量归一化 && 残差网络
基于此前对于CNN的介绍
- Fundamentals of Convolutional Neural Networks
- LeNet && ModernCNN
就深层次 CNN 的结构进一步探讨归一化和残差网络。
批量归一化（BatchNormalization）
让网络训练归一化变得更加容易，本质是一种对数据的标准化处理
分类
- 对输入的标准化（浅层模型）
处理后的任意一个特征在数据集中所有样本上的均值为0、标准差为1。 标准化处理输入数据使各个特征的分布相近
- 批量归一化（深度模型）随着模型参数的迭代更新，靠近输出层的数据剧烈变化
利用小批量上的均值和标准差，不断调整神经网络中间输出，从而使整个神经网络在各层的中间输出的数值更稳定。
应用
- 对全连接层做批量归一化
位置：全连接层中的仿射变换和激活函数之间。
全连接：
x = W u + b o u t p u t = ϕ ( x ) \boldsymbol{x} = \boldsymbol{W\boldsymbol{u} + \boldsymbol{b}} \\ output =\phi(\boldsymbol{x}) x=Wu+boutput=ϕ(x)
输入是u，经过仿射变化得到x，经过激活函数得到output，size=(batch_size，输出神经元的个数)
批量归一化：
o u t p u t = ϕ ( BN ( x ) ) output=\phi(\text{BN}(\boldsymbol{x})) output=ϕ(BN(x))
y ( i ) = BN ( x ( i ) ) \boldsymbol{y}^{(i)} = \text{BN}(\boldsymbol{x}^{(i)}) y(i)=BN(x(i))
μ B ← 1 m ∑ i = 1 m x ( i ) , \boldsymbol{\mu}_\mathcal{B} \leftarrow \frac{1}{m}\sum_{i = 1}^{m} \boldsymbol{x}^{(i)}, μB←m1i=1∑mx(i),
σ B 2 ← 1 m ∑ i = 1 m ( x ( i ) − μ B ) 2 , \boldsymbol{\sigma}_\mathcal{B}^2 \leftarrow \frac{1}{m} \sum_{i=1}^{m}(\boldsymbol{x}^{(i)} - \boldsymbol{\mu}_\mathcal{B})^2, σB2←m1i=1∑m(x(i)−μB)2,
x ^ ( i ) ← x ( i ) − μ B σ B 2 + ϵ , \hat{\boldsymbol{x}}^{(i)} \leftarrow \frac{\boldsymbol{x}^{(i)} - \boldsymbol{\mu}_\mathcal{B}}{\sqrt{\boldsymbol{\sigma}_\mathcal{B}^2 + \epsilon}}, x^(i)←σB2+ϵx(i)−μB,
标 准 化 处 理 标准化处理 标准化处理
这⾥ϵ > 0是个很小的常数，保证分母大于0
y ( i ) ← γ ⊙ x ^ ( i ) + β . {\boldsymbol{y}}^{(i)} \leftarrow \boldsymbol{\gamma} \odot \hat{\boldsymbol{x}}^{(i)} + \boldsymbol{\beta}. y(i)←γ⊙x^(i)+β.
引入可学习参数：拉伸参数γ和偏移参数β。若 γ = σ B 2 + ϵ \boldsymbol{\gamma} = \sqrt{\boldsymbol{\sigma}_\mathcal{B}^2 + \epsilon} γ=σB2+ϵ和 β = μ B \boldsymbol{\beta} = \boldsymbol{\mu}_\mathcal{B} β=μB，批量归一化无效。
- 对卷积层做批量归⼀化
位置：卷积计算之后、应⽤激活函数之前
如果卷积计算输出多个通道，我们需要对这些通道的输出分别做批量归一化，且每个通道都拥有独立的拉伸和偏移参数。
计算：对单通道， b a t c h s i z e = m , 卷 积 计 算 输 出 = p × q batchsize = m,卷积计算输出 = p \times q batchsize=m,卷积计算输出=p×q
对该通道中 m × p × q m\times p\times q m×p×q 个元素同时做批量归一化,使用相同的均值和方差。
- 预测时的批量归⼀化
训练：以 batch 为单位, 对每个 batch 计算均值和方差。
预测：用移动平均估算整个训练数据集的样本均值和方差。
实现
batch_norm function
import time
import torch
from torch import nn, optim
import torch.nn.functional as F
import torchvision
import sys
sys.path.append("path to file storge d2lzh1981")
import d2lzh1981 as d2l
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# 放到了BatchNorm类中，使用时直接调用forward函数，此函数将成为cell函数
def batch_norm(is_training, X, gamma, beta, moving_mean, moving_var, eps, momentum):
# 判断当前模式是训练模式还是预测模式
if not is_training:
# 如果是在预测模式下，直接使用传入的移动平均所得的均值和方差
X_hat = (X - moving_mean) / torch.sqrt(moving_var + eps)
else:
assert len(X.shape) in (2, 4)
if len(X.shape) == 2:
# 使用全连接层的情况，计算特征维上的均值和方差
mean = X.mean(dim=0) #是d维的值
var = ((X - mean) ** 2).mean(dim=0)
else:
# 使用二维卷积层的情况，计算通道维上（axis=1）的均值和方差。这里我们需要保持
# X的形状以便后面可以做广播运算
mean = X.mean(dim=0, keepdim=True).mean(dim=2, keepdim=True).mean(dim=3, keepdim=True) # c维的值，通道有几个mean就有几个
var = ((X - mean) ** 2).mean(dim=0, keepdim=True).mean(dim=2, keepdim=True).mean(dim=3, keepdim=True)
# 训练模式下用当前的均值和方差做标准化
X_hat = (X - mean) / torch.sqrt(var + eps)
# 更新移动平均的均值和方差
# momentum 是一个超参数
moving_mean = momentum * moving_mean + (1.0 - momentum) * mean
moving_var = momentum * moving_var + (1.0 - momentum) * var
Y = gamma * X_hat + beta # 拉伸和偏移
return Y, moving_mean, moving_var
batch_norm class
# 在Batch Norm函数的基础上定义此类，作用是维护学习参数和超参数
class BatchNorm(nn.Module):
def __init__(self, num_features, num_dims):
super(BatchNorm, self).__init__()
if num_dims == 2:
shape = (1, num_features) #全连接层输出神经元
# num_features代表输出神经元的个数
else:
shape = (1, num_features, 1, 1) #通道数
# 参与求梯度和迭代的拉伸和偏移参数，分别初始化成0和1
self.gamma = nn.Parameter(torch.ones(shape))
self.beta = nn.Parameter(torch.zeros(shape))
# 不参与求梯度和迭代的变量，全在内存上初始化成0
self.moving_mean = torch.zeros(shape)
self.moving_var = torch.zeros(shape)
def forward(self, X):
# 如果X不在内存上，将moving_mean和moving_var复制到X所在显存上
if self.moving_mean.device != X.device:
self.moving_mean = self.moving_mean.to(X.device)
self.moving_var = self.moving_var.to(X.device)
# 保存更新过的moving_mean和moving_var, Module实例的traning属性默认为true, 调用.eval()后设成false
Y, self.moving_mean, self.moving_var = batch_norm(self.training,
X, self.gamma, self.beta, self.moving_mean,
self.moving_var, eps=1e-5, momentum=0.9)
return Y
基于LeNet的应用
net = nn.Sequential(
nn.Conv2d(1, 6, 5), # in_channels, out_channels, kernel_size
# 直接作为一个参数加到LeNet中就行
BatchNorm(6, num_dims=4),
nn.Sigmoid(),
nn.MaxPool2d(2, 2), # kernel_size, stride
# 体现了仿射函数之后激活函数之前的结构
nn.Conv2d(6, 16, 5),
BatchNorm(16, num_dims=4),
nn.Sigmoid(),
nn.MaxPool2d(2, 2),
d2l.FlattenLayer(),
nn.Linear(16*4*4, 120),
BatchNorm(120, num_dims=2),
nn.Sigmoid(),
nn.Linear(120, 84),
BatchNorm(84, num_dims=2),
nn.Sigmoid(),
nn.Linear(84, 10)
)
print(net)
load data
#batch_size = 256
##cpu要调小batchsize
batch_size=16
def load_data_fashion_mnist(batch_size, resize=None, root='/home/kesci/input/FashionMNIST2065'):
"""Download the fashion mnist dataset and then load into memory."""
trans = []
if resize:
trans.append(torchvision.transforms.Resize(size=resize))
trans.append(torchvision.transforms.ToTensor())
transform = torchvision.transforms.Compose(trans)
mnist_train = torchvision.datasets.FashionMNIST(root=root, train=True, download=True, transform=transform)
mnist_test = torchvision.datasets.FashionMNIST(root=root, train=False, download=True, transform=transform)
train_iter = torch.utils.data.DataLoader(mnist_train, batch_size=batch_size, shuffle=True, num_workers=2)
test_iter = torch.utils.data.DataLoader(mnist_test, batch_size=batch_size, shuffle=False, num_workers=2)
return train_iter, test_iter
train_iter, test_iter = load_data_fashion_mnist(batch_size)
train and test
lr, num_epochs = 0.001, 5
optimizer = torch.optim.Adam(net.parameters(), lr=lr)
d2l.train_ch5(net, train_iter, test_iter, batch_size, optimizer, device, num_epochs)
简化模型
nn中有内置的BatchNorm2d（卷积层）和BatchNorm1d（全连接层）
'''
在自己应用时不需要写class和function
'''
net = nn.Sequential(
nn.Conv2d(1, 6, 5), # in_channels, out_channels, kernel_size
nn.BatchNorm2d(6),
nn.Sigmoid(),
nn.MaxPool2d(2, 2), # kernel_size, stride
nn.Conv2d(6, 16, 5),
nn.BatchNorm2d(16),
nn.Sigmoid(),
nn.MaxPool2d(2, 2),
d2l.FlattenLayer(),
nn.Linear(16*4*4, 120),
nn.BatchNorm1d(120),
nn.Sigmoid(),
nn.Linear(120, 84),
nn.BatchNorm1d(84),
nn.Sigmoid(),
nn.Linear(84, 10)
)
optimizer = torch.optim.Adam(net.parameters(), lr=lr)
d2l.train_ch5(net, train_iter, test_iter, batch_size, optimizer, device, num_epochs)
残差网络（ResNet）
深度学习的问题
深度CNN网络达到一定深度后再一味地增加层数并不能带来进一步地分类性能提高，反而会招致网络收敛变得更慢，准确率也变得更差。
残差块（Residual Block）
恒等映射：
- 左边： f ( x ) = x f(x)=x f(x)=x
- 右边： f ( x ) − x = 0 f(x)-x=0 f(x)−x=0 （易于捕捉恒等映射的细微波动 ; 易于优化）
神 经 网 络 普 通 层 ( l e f t ) 残 差 网 络 ( r i g h t ) 神经网络普通层(left)残差网络(right) 神经网络普通层(left)残差网络(right)
在残差块中，输⼊可通过跨层的数据线路更快 地向前传播。
残差块实现
class Residual(nn.Module):
#可以设定输出通道数、是否使用额外的1x1卷积层来修改通道数以及卷积层的步幅。
def __init__(self, in_channels, out_channels, use_1x1conv=False, stride=1):
super(Residual, self).__init__()
self.conv1 = nn.Conv2d(in_channels, out_channels, kernel_size=3, padding=1, stride=stride)
self.conv2 = nn.Conv2d(out_channels, out_channels, kernel_size=3, padding=1)
if use_1x1conv:
self.conv3 = nn.Conv2d(in_channels, out_channels, kernel_size=1, stride=stride)
else:
self.conv3 = None
self.bn1 = nn.BatchNorm2d(out_channels)
self.bn2 = nn.BatchNorm2d(out_channels)
def forward(self, X):
Y = F.relu(self.bn1(self.conv1(X)))
Y = self.bn2(self.conv2(Y))
if self.conv3:
X = self.conv3(X)
return F.relu(Y + X)
是否需要 1 × 1 1\times1 1×1 卷积层
# 不需要使用1*1卷积层 输入和输出相同
blk = Residual(3, 3)
X = torch.rand((4, 3, 6, 6))
blk(X).shape # torch.Size([4, 3, 6, 6])
#需要使用
blk = Residual(3, 6, use_1x1conv=True, stride=2)
blk(X).shape # torch.Size([4, 6, 3, 3])
ResNet模型
简化实现
net = nn.Sequential(
nn.Conv2d(1, 64, kernel_size=7, stride=2, padding=3),
nn.BatchNorm2d(64),
nn.ReLU(),
nn.MaxPool2d(kernel_size=3, stride=2, padding=1))
残差块
def resnet_block(in_channels, out_channels, num_residuals, first_block=False):
if first_block:
assert in_channels == out_channels # 第一个模块的通道数同输入通道数一致
blk = []
for i in range(num_residuals):
if i == 0 and not first_block:
# 把in_channels放缩到out_channels的个数
blk.append(Residual(in_channels, out_channels, use_1x1conv=True, stride=2))
else:
# 保证输入和输出都是out_channels
blk.append(Residual(out_channels, out_channels))
return nn.Sequential(*blk)
# 把四个残差block放到net里
net.add_module("resnet_block1", resnet_block(64, 64, 2, first_block=True))
net.add_module("resnet_block2", resnet_block(64, 128, 2))
net.add_module("resnet_block3", resnet_block(128, 256, 2))
net.add_module("resnet_block4", resnet_block(256, 512, 2))
全局平均池化
net.add_module("global_avg_pool", d2l.GlobalAvgPool2d()) # GlobalAvgPool2d的输出: (Batch, 512, 1, 1)
net.add_module("fc", nn.Sequential(d2l.FlattenLayer(), nn.Linear(512, 10)))
train and test
lr, num_epochs = 0.001, 5
optimizer = torch.optim.Adam(net.parameters(), lr=lr)
d2l.train_ch5(net, train_iter, test_iter, batch_size, optimizer, device, num_epochs)
ResNet 的引申设计
稠密连接网络（DenseNet）
特征：concat 连接
主要构建模块：
- 稠密块（dense block）： 定义了输入和输出是如何连结的。
- 过渡层（transition layer）：用来控制通道数，使之不过大。
稠密块
- 输出通道数=输入通道数+卷积层个数*卷积输出通道数
def conv_block(in_channels, out_channels):
blk = nn.Sequential(nn.BatchNorm2d(in_channels),
nn.ReLU(),
nn.Conv2d(in_channels, out_channels, kernel_size=3, padding=1))
return blk
class DenseBlock(nn.Module):
# num_convs:用了几个卷积层
# in_channels代表全部输入，但是out_channels不代表全部输出
def __init__(self, num_convs, in_channels, out_channels):
super(DenseBlock, self).__init__()
net = []
for i in range(num_convs):
# 卷积层输入的通道数
in_c = in_channels + i * out_channels
net.append(conv_block(in_c, out_channels))
self.net = nn.ModuleList(net)
self.out_channels = in_channels + num_convs * out_channels # 计算输出通道数
def forward(self, X):
for blk in self.net:
Y = blk(X)
# concat连接
X = torch.cat((X, Y), dim=1) # 在通道维上将输入和输出连结
return X
blk = DenseBlock(2, 3, 10)
X = torch.rand(4, 3, 8, 8)
Y = blk(X)
Y.shape # torch.Size([4, 23, 8, 8]) 3+2*10
过渡层
- 1 × 1 1\times1 1×1卷积层：来减小通道数
- 步幅为2的平均池化层：减半高和宽
def transition_block(in_channels, out_channels):
blk = nn.Sequential(
nn.BatchNorm2d(in_channels),
nn.ReLU(),
nn.Conv2d(in_channels, out_channels, kernel_size=1),
nn.AvgPool2d(kernel_size=2, stride=2))
return blk
blk = transition_block(23, 10)
blk(Y).shape # torch.Size([4, 10, 4, 4])
DenseNet模型
net = nn.Sequential(
# 宽高减半
nn.Conv2d(1, 64, kernel_size=7, stride=2, padding=3),
nn.BatchNorm2d(64),
nn.ReLU(),
# 宽高减半
nn.MaxPool2d(kernel_size=3, stride=2, padding=1))
num_channels, growth_rate = 64, 32 # num_channels为当前的通道数
num_convs_in_dense_blocks = [4, 4, 4, 4]
for i, num_convs in enumerate(num_convs_in_dense_blocks):
DB = DenseBlock(num_convs, num_channels, growth_rate)
net.add_module("DenseBlosk_%d" % i, DB)
# 上一个稠密块的输出通道数
num_channels = DB.out_channels
# 在稠密块之间加入通道数减半的过渡层
if i != len(num_convs_in_dense_blocks) - 1:
net.add_module("transition_block_%d" % i, transition_block(num_channels, num_channels // 2))
num_channels = num_channels // 2
net.add_module("BN", nn.BatchNorm2d(num_channels))
net.add_module("relu", nn.ReLU())
net.add_module("global_avg_pool", d2l.GlobalAvgPool2d()) # GlobalAvgPool2d的输出: (Batch, num_channels, 1, 1)
net.add_module("fc", nn.Sequential(d2l.FlattenLayer(), nn.Linear(num_channels, 10)))
# 测试
X = torch.rand((1, 1, 96, 96))
for name, layer in net.named_children():
X = layer(X)
print(name, ' output shape:\t', X.shape)
#batch_size = 256
batch_size=16
# 如出现“out of memory”的报错信息，可减小batch_size或resize
train_iter, test_iter =load_data_fashion_mnist(batch_size, resize=96)
lr, num_epochs = 0.001, 5
optimizer = torch.optim.Adam(net.parameters(), lr=lr)
d2l.train_ch5(net, train_iter, test_iter, batch_size, optimizer, device, num_epochs)