import numpy as np
import matplotlib.pyplot as plt
from torch import  nn,optim
from torch.autograd import Variable
import torch

x_data = np.linspace(-2,2,200)[:,np.newaxis]

noise = np.random.normal(0,0.2,x_data.shape)
y_data = np.square(x_data) + noise

#plt.scatter(x_data,y_data)

#plt.show()

x_data = x_data.reshape(-1,1)
y_data = y_data.reshape(-1,1)

#把numpy转换成tensor
x_data = torch.FloatTensor(x_data)
y_data = torch.FloatTensor(y_data)

inputs = Variable(x_data)
target = Variable(y_data)

#构造神经网络模型
#一般把网络中具有可学习参数的层放在__init__()中
class LinearRegression(nn.Module):
    #定义网络结构
    def __init__(self):
        #初始化nn.Module
        super(LinearRegression,self).__init__()
        #1 - 10 - 1
        self.fc1 = nn.Linear(1,10)    #输入一个神经元，输出一个神经元
        self.tanh = nn.Tanh()
        self.fc2 = nn.Linear(10,1)

    #定义网络计算
    def forward(self,x):
        x = self.fc1(x)
        x = self.tanh(x)
        x = self.fc2(x)
        return x

#定义模型
model = LinearRegression()
#定义代价函数
mse_loss = nn.MSELoss()     #均方差
#定义优化器
optimizer = optim.SGD(model.parameters(),lr=0.3)

for name,parameters in model.named_parameters():
    print('name:{},param{}'.format(name,parameters))


for i in range(2001):
    out = model(inputs)
    #计算loss
    loss = mse_loss(out,target)
    #梯度清0
    optimizer.zero_grad()
    #计算梯度
    loss.backward()
    #修改权值
    optimizer.step()

    if i % 200 == 0:
        print(i, loss.item())

y_pred = model(inputs)
plt.scatter(x_data, y_data)
plt.plot(x_data, y_pred.data.numpy(), 'r-', lw=3)
plt.show()


输出的结果

python 非线性回归 pytorch 非线性回归_神经网络模型

name:fc1.weight,paramParameter containing:
tensor([[-0.8655],
        [ 0.5277],
        [ 0.4221],
        [-0.2834],
        [ 0.9853],
        [-0.0767],
        [-0.8628],
        [-0.0492],
        [ 0.6841],
        [ 0.2099]], requires_grad=True)
name:fc1.bias,paramParameter containing:
tensor([-0.7933,  0.5301,  0.0711, -0.8077,  0.5706, -0.4823,  0.9381,  0.7663,
        -0.3518, -0.7270], requires_grad=True)
name:fc2.weight,paramParameter containing:
tensor([[-0.1524, -0.0411,  0.0731, -0.3065, -0.1577, -0.0627, -0.0596, -0.2775,
          0.0461,  0.1572]], requires_grad=True)
name:fc2.bias,paramParameter containing:
tensor([-0.1086], requires_grad=True)
0 3.856107711791992
200 0.2011813372373581
400 0.09159507602453232
600 0.06196943297982216
800 0.05489378049969673
1000 0.05240625888109207
1200 0.05162370577454567
1400 0.05135105550289154
1600 0.05146040394902229
1800 0.051464591175317764
2000 0.05113458260893822

Process finished with exit code 0