分类问题
感知机:二分类问题
线性回归是输出实数,感知机是输出离散类
Softmax 可实现多分类,感知机 实现二分类
收敛定理(停止条件):所有数据分类正确,且有一定余量
感知机不能拟合XOR问题,只能产生线性分割面
多层感知机
首先解决XOR问题==>做多次感知分类
激活函数 要使用非线性函数
- 将 x 的值投影到一个0和1的开区间中
- sigmoid实际上是阶跃函数的温和版
y = torch.sigmoid(x)
d2l.plot(x.detach(), y.detach(), 'x', 'sigmoid(x)', figsize=(5, 2.5))
- 和sigmoid很像,区别在于它是将输入投影到-1到1的区间内
- -2是为了方便求导
y = torch.tanh(x)
d2l.plot(x.detach(), y.detach(), 'x', 'tanh(x)', figsize=(5, 2.5))
- 最常用
- 不用做指数运算
x = torch.arange(-8.0, 8.0, 0.1, requires_grad=True)
y = torch.relu(x)
d2l.plot(x.detach(), y.detach(), 'x', 'relu(x)', figsize=(5, 2.5))
- softmax就是将所有的输入映射到0和1的区间之内,并且所有输出的值加起来等于1,从而转变成概率
1.做线性
2.单隐藏层
3.多隐藏层
简洁代码实现
1 导入包
import torch
from torch import nn
from d2l import torch as d2l
2 创建网络
这里784是一个数据的输入,就是说这个网络输入的点有784个。256表示的是隐藏层的单元个数。10表示的是最终的输出是一个长为10的向量。
nn.Sequential()
net = nn.Sequential(nn.Flatten(),
nn.Linear(784, 256),
nn.ReLU(),
nn.Linear(256, 10)) #可以理解为网络是由这4部分组成的
def init_weights(m):
if type(m) == nn.Linear:
nn.init.normal_(m.weight, std=0.01)
net.apply(init_weights);
3 训练
batch_size, lr, num_epochs = 256, 0.1, 10
loss = nn.CrossEntropyLoss()
trainer = torch.optim.SGD(net.parameters(), lr=lr)
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size)
d2l.train_ch3(net, train_iter, test_iter, loss, num_epochs, trainer)
从零开始代码实现
1 导入包
import torch
from torch import nn
from d2l import torch as d2l
batch_size = 256 #批量大小为256
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size) #导入训练和测试数据集
2 参数w和b初始化
输入和输出个数参考softmax函数是实现根据模型固定了的。
Parameter torch.randn
num_inputs, num_outputs, num_hiddens = 784, 10, 256 #输入、输出、隐藏层单元 个数。
W1 = nn.Parameter(torch.randn(
num_inputs, num_hiddens, requires_grad=True) * 0.01) #w初始化成一个随机的784*256的tensor
b1 = nn.Parameter(torch.zeros(num_hiddens, requires_grad=True)) #b初始化一个
W2 = nn.Parameter(torch.randn(
num_hiddens, num_outputs, requires_grad=True) * 0.01)
b2 = nn.Parameter(torch.zeros(num_outputs, requires_grad=True))
params = [W1, b1, W2, b2]
3 ReLu函数
def relu(X):
a = torch.zeros_like(X)
return torch.max(X, a)
4 定义模型
即前向传播的计算
nn.CrossEntropyLoss()
def net(X):
X = X.reshape((-1, num_inputs))
H = relu(X@W1 + b1) # 这里“@”代表矩阵乘法
return (H@W2 + b2)
loss = nn.CrossEntropyLoss() #定义loss损失函数为交叉熵损失函数
5 训练
num_epochs, lr = 10, 0.1 #循环10次,学习率为0.1
updater = torch.optim.SGD(params, lr=lr) #随机梯度下降算法
d2l.train_ch3(net, train_iter, test_iter, loss, num_epochs, updater) #画图
6 预测
d2l.predict_ch3(net, test_iter)
完整代码:
import torch
from torch import nn
from d2l import torch as d2l
batch_size = 256
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size)
'''
# 实现一个具有单隐藏层的多层感知机,它包含256个隐藏单元
num_inputs, num_outputs, num_hiddens = 784, 10, 256
W1 = nn.Parameter(torch.randn(num_inputs, num_hiddens, requires_grad=True))
print(f'W1.shape: {W1.shape}')
b1 = nn.Parameter(torch.zeros(num_hiddens, requires_grad=True))
print(f'b1.shape: {b1.shape}')
W2 = nn.Parameter(torch.randn(num_hiddens, num_outputs, requires_grad=True))
b2 = nn.Parameter(torch.zeros(num_outputs, requires_grad=True))
params = [W1, b1, W2, b2]
# 实现ReLU激活函数
def relu(X):
a = torch.zeros_like(X)
return torch.max(X, a)
# 实现我们的模型
def net(X):
X = X.reshape((-1, num_inputs))
print(f'X.shape: {X.shape}')
H = relu(X @ W1 + b1)
return (H @ W2 + b2)
loss = nn.CrossEntropyLoss()
num_epochs, lr = 10, 0.1
updater = torch.optim.SGD(params, lr=lr)
d2l.train_ch3(net, train_iter, test_iter, loss, num_epochs, updater)
'''
# 简洁实现
# 输入是3D的,要Flatten成2D
net = nn.Sequential(nn.Flatten(), nn.Linear(784, 256), nn.ReLU(), nn.Linear(256, 10))
def init_weights(m):
if type(m) == nn.Linear:
nn.init.normal_(m.weight, std=0.01)
net.apply(init_weights)
batch_size, lr, num_epochs = 256, 0.1, 10
loss = nn.CrossEntropyLoss()
trainer = torch.optim.SGD(net.parameters(), lr=lr)
d2l.train_ch3(net, train_iter, test_iter, loss, num_epochs, trainer)
pycharm版本:
import torch
from torch import nn
from d2l import torch as d2l
net = nn.Sequential(
nn.Flatten(), # flatten: 展平层,将输入数据展平,这里只保留第0维度
nn.Linear(784, 256),
nn.ReLU(),
nn.Linear(256, 10)
)
def init_weights(m): # m: 当前layer
if type(m) == nn.Linear:
nn.init.normal_(m.weight, std=0.01)
# m.weight 默认为0,这一行的意思就是说把当前layer层的参数初始化成均值为0,标准差为0.01
net.apply(init_weights) # 每一层的权值都进行初始化
batch_size, lr, num_epochs = 256, 0.1, 10
loss = nn.CrossEntropyLoss(reduction='none') # 交叉熵损失函数
trainer = torch.optim.SGD(net.parameters(), lr=lr) # 优化器选择随机梯度下降
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size)
d2l.train_ch3(net, train_iter, test_iter, loss, num_epochs, trainer)
d2l.plt.show()