[Master Liu Er] Pytorch deep learning practice (1): code and drawing implementation of linear models y=w*x and y=w*x+b

Learning video: Master Liu pytorch1-linear model

Question description

1. y=w*x

Use the linear model y = w*x to calculate the loss value of the model and use matplotlib to give the cost function image.

2. y=w*x+b

Use the linear model y = w*x+b to give the cost function image, and use 3D graphics to draw the image.

code

Question 1: y = w*x

import numpy as np
import matplotlib.pyplot as plt

# 训练数据
x_data = [1.0,2.0,3.0]
y_data = [2.0,4.0,6.0]

# 前向传播计算预测值
def forward(x):
    return x*w

# 计算损失值的平方
def loss(x,y):
    y_pred = forward(x)
    return (y-y_pred)*(y-y_pred)

# 记录画图的x值和y值
w_list = []
loss_mse_list = []
# 计算数据集中数据个数
n = len(x_data)
for w in np.arange(0.0,4.1,0.1):
    w_list.append(w)
    print("w=",w)
    # 存储每一个数据对的loss值的和
    loss_sum = 0
    for x,y in zip(x_data,y_data):
        y_pred = forward(x)
        # 一个数据对的loss值
        loss1 = loss(x,y)
        loss_sum += loss1
        print('\t',x,y,y_pred,loss1)
        # 对loss和求平均
    loss_mse_list.append(loss_sum/n)
    print('MSE=',loss_sum/n)

plt.plot(w_list,loss_mse_list)
plt.xlabel("w")
plt.ylabel("mse loss")
plt.show()

Why should process information be printed?

Record logs to prevent loss of training data when the program stops abnormally during training.

Question 2: y = wx+b

Drawing:matplotlib-3d drawing method

import numpy as np
import matplotlib.pyplot as plt

# 训练数据
x_data = [1.0,2.0,3.0]
y_data = [3.0,5.0,7.0]

# 前向传播计算预测值
def forward(x):
    return x*w+b

# 计算损失值的平方
def loss(x,y):
    y_pred = forward(x)
    return (y-y_pred)*(y-y_pred)

# 记录画图的x值和y值
w_list = np.arange(0.0,4.1,0.1)
b_list = np.arange(-2.0,2.1,0.1)
loss_mse_list = []
# 计算数据集中数据个数
n = len(x_data)

for w in w_list:
    for b in b_list:
        l_sum = 0
        for x,y in zip(x_data,y_data):
            l_sum+=loss(x,y)
        loss_mse_list.append(l_sum/n)

w_list, b_list=np.meshgrid(w_list, b_list)
mse_list=np.array(loss_mse_list).reshape(len(w_list),-1)

fig = plt.figure()  # 创建图片
sub = fig.add_subplot(111, projection='3d')  # 添加子图，
surf = sub.plot_surface(w_list, b_list, mse_list, cmap=plt.cm.brg)  # 绘制曲面,并设置颜色cmap
cb = fig.colorbar(surf, shrink=0.8, aspect=15)  # 设置颜色棒

sub.set_xlabel(r"$w$")
sub.set_ylabel(r"$b$")
sub.set_zlabel(r"$loss$")
plt.show()

drawing tools

wisdom ：wisdom usage