Station B Liu Erden - linear regression and gradient descent

List of articles

《PyTorch Deep learning practice 》 Complete the collection -B Stand up, Mr. Liu er

Pytorch Details of code attention , Easy to hit the wrong place

B Stand up, Mr. Liu er - Back propagation Lecture 3

B Stand up, Mr. Liu er - Linear regression Pytorch Lecture 4

---

---

Preface

Lecture 3

---

The graph of self written code is a straight line , reason ： Learning rate alpha Set too large , from 0.1 Set to 0.005 It will become a curve

One 、 Code

import matplotlib.pyplot as plt
import  torch
import numpy as np
import  csv
import os
x_data = [1.0, 2.0, 3.0]
y_data = [2.0, 4.0, 6.0]

def forward(x,w):
    return  x*w

def gradient(xs, ys,w):
    grad = 0
    for(x,y) in zip(xs,ys):
        grad += (x*w-y)*x
    grad = 2*grad/len(xs)
    return grad

def loss(xs, ys, w):
    los = 0
    for(x,y) in zip(xs,ys):
        los += (x*w-y)*(x*w-y)
    los = los/len(xs)
    return  los

w_list = []
mse_list = []
alpha = 0.1
w = 1.0
ite = 100
for i in range(ite):
    los = loss(x_data, y_data, w)
    mse_list.append(los / 3.0)
    w_list.append(w)
    w = w - alpha*gradient(x_data, y_data, w)

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

Above learning rate 0.1

Upper figure ： Learning rate 0.005

---