批量归一化
-
批量归一化所考虑的问题是在学习底部层的时候避免变化顶部层?
-
损失出现在最后,后面的层训练较快。
-
数据在最底部
- 底部的层训练较慢
- 底部层一变化,所有都得跟着变
- 最后的那些层需要重新学习多次
- 导致收敛变慢
-
核心想法:固定小批量里面的均值和方差:$\mu_B = \frac{1}{\left | B \right |} \sum_{i\in B}x_i and \sigma^2_B=\frac{1}{\left | B \right |}\sum_{i\in B}(x_i - \mu_B)^2 + \epsilon $。这里B就是我们的所有小批量的数据。这里可以假设x是一个向量便于拓展到神经网络。方差中加上一个很小的数,放止其变为0
-
批量归一化其实就是在做的事情就是减去均值除以方差乘以 γ \gamma γ在加上 β \beta β,这里 γ , β \gamma, \beta γ,β是学习的参数,其余的参数都是根据数据来的,但是在这里会限定住这两个值不要变化的过于猛烈。
批量归一化层
- 可学习的参数 γ \gamma γ和 β \beta β
- 作用在
- 全连接层和卷积层输出上,激活函数前
- 全连接层和卷积层输入上
- 对全连接层,作用在特征维
- 对卷积层,作用在通道维
批量归一化在做什么?
总结
从零开始实现批量归一化
import torch
from torch import nn
from d2l import torch as d2l
# X是输入,gamma和beta是我们要学的,moving_mean和moving_var全局的一个均值和方差,eps避免出0,momentum用来更新这个moving_mean和moving_var的东西
def batch_norm(X, gamma, beta, moving_mean, moving_var, eps, momentum):
# 不要算梯度,不是在trainning
if not torch.is_grad_enabled():
X_hat = (X - moving_mean) / torch.sqrt(moving_var + eps)
else:
assert len(X.shape) in (2, 4)
# 这里输入是二维的,也就是全连接
if len(X.shape) == 2:
mean = X.mean(dim=0)
var = ((X - mean) ** 2).mean(dim=0) # 按行求均值,也就是每一列求出一个均值出来
# 下面是2d卷积的情况
else:
mean = X.mean(dim=(0, 2, 3), keepdim=True) # 这里的意思对于通道维,将其他的所有的东西都拿出来求平均值。得出(1, n, 1, 1)
var = ((X - mean) ** 2).mean(dim=(0, 2, 3), keepdim=True)
var = ((X - mean) ** 2).mean(dim=(0, 2, 3), keepdim=True)
X_hat = (X - mean) / torch.sqrt(var + eps) # 这个是当前的小批量里面的mean和方差
moving_mean = momentum * moving_mean + (1.0 - momentum) * mean # 更新全局的mean
moving_var = momentum * moving_var + (1.0 - momentum) * var
Y = gamma * X_hat + beta
return Y, moving_mean.data, moving_var.data
# 创建一个正确的BatchNorm图层
class BatchNorm(nn.Module):
def __init__(self, num_features, num_dims):
super().__init__()
if num_dims == 2:
shape = (1, num_features)
else:
shape = (1, num_features, 1, 1)
self.gamma = nn.Parameter(torch.ones(shape))
self.beta = nn.Parameter(torch.zeros(shape))
self.moving_mean = torch.zeros(shape)
self.moving_var = torch.ones(shape)
def forward(self, X):
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)
Y, self.moving_mean, self.moving_var = batch_norm(
X, self.gamma, self.beta, self.moving_mean, self.moving_var,
eps=1e-5, momentum=0.9)
return Y
# 应用BatchNorm于LeNet模型
net = nn.Sequential(nn.Conv2d(1, 6, kernel_size=5), BatchNorm(6, num_dims=4),
nn.Sigmoid(), nn.MaxPool2d(kernel_size=2, stride=2),
nn.Conv2d(6, 16,kernel_size=5), BatchNorm(16, num_dims=4),
nn.Sigmoid(), nn.MaxPool2d(kernel_size=2, stride=2),
nn.Flatten(), 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))
lr, num_epochs, batch_size = 1.0, 10, 256
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size)
d2l.train_ch6(net, train_iter, test_iter, num_epochs, lr, d2l.try_gpu())
loss 0.246, train acc 0.910, test acc 0.801
4279.8 examples/sec on cpu
# 简洁实现
net = nn.Sequential(nn.Conv2d(1, 6, kernel_size=5), nn.BatchNorm2d(6),
nn.Sigmoid(), nn.MaxPool2d(kernel_size=2, stride=2),
nn.Conv2d(6, 16, kernel_size=5), nn.BatchNorm2d(16),
nn.Sigmoid(), nn.MaxPool2d(kernel_size=2, stride=2),
nn.Flatten(), nn.Linear(256, 120), nn.BatchNorm1d(120),
nn.Sigmoid(), nn.Linear(120, 84), nn.BatchNorm1d(84),
nn.Sigmoid(), nn.Linear(84, 10))
# 使用相同的超参数来训练模型
d2l.train_ch6(net, train_iter, test_iter, num_epochs, lr, d2l.try_gpu())
loss 0.246, train acc 0.910, test acc 0.856
6917.9 examples/sec on cpu