# 线性方程求解—损失函数—梯度下降

``````%matplotlib inline
import pandas as pd
import matplotlib.pylab as plt
``````# Normalize the data 归一化值 (x - mean) / (std)
pga.distance = (pga.distance - pga.distance.mean())/pga.distance.std()
pga.accuracy = (pga.accuracy - pga.accuracy.mean())/pga.accuracy.std()``````
``````#将归一化的散点数据画出来
plt.scatter(pga.distance, pga.accuracy) #画散点图
plt.xlabel('normalized distance')
plt.ylabel('normalized accuracy')
plt.show()``````
``````from sklearn.linear_model import LinearRegression
import numpy as np
lm = LinearRegression()
lm.fit(pga.distance[:, np.newaxis],pga.accuracy)  #newaxis ,给数据增加维数。参见https://blog.csdn.net/xtingjie/article/details/72510834
thetal = lm.coef_
print(thetal)``````

-0.6075988227151216

``````# 单变量 代价函数
def cost(theta0, theta1, x, y):
# Initialize cost
J = 0
# The number of observations
m = len(x)
# Loop through each observation
# 通过每次观察进行循环
for i in range(m):
# Compute the hypothesis
# 计算假设
h = theta1 * x[i] + theta0
J += (h - y[i])**2
# Average and normalize cost
J /= (2*m)
return J

# The cost for theta0=0 and theta1=1 针对单个theta0,theta1的值，计算得到损失值
print(cost(0, 1, pga.distance, pga.accuracy))``````

1.599438422599817

``````theta0 = 100  #赋theta0初值
theta1s = np.linspace(-3,2,100)  #生成-3到2间的100个均匀间隔数字。
costs = []
for theta1 in theta1s:   #计算theta0=100,theta1从-3到2间时损失值
costs.append(cost(theta0, theta1, pga.distance, pga.accuracy))

plt.plot(theta1s, costs)
plt.show()``````
``````#对于三维数据的计算
import numpy as np
from mpl_toolkits.mplot3d import Axes3D

# Example of a Surface Plot using Matplotlib
# Create x an y variables
x = np.linspace(-10,10,100)
y = np.linspace(-10,10,100)

# We must create variables to represent each possible pair of points in x and y
# ie. (-10, 10), (-10, -9.8), ... (0, 0), ... ,(10, 9.8), (10,9.8)
# x and y need to be transformed to 100x100 matrices to represent these coordinates
# np.meshgrid will build a coordinate matrices of x and y
X, Y = np.meshgrid(x,y)
#print(X[:5,:5],"\n",Y[:5,:5])

# Compute a 3D parabola
Z = X**2 + Y**2

# Open a figure to place the plot on
fig = plt.figure()
# Initialize 3D plot
ax = fig.gca(projection='3d')
# Plot the surface
ax.plot_surface(X=X,Y=Y,Z=Z)

plt.show()

# Use these for your excerise
theta0s = np.linspace(-2,2,100)
theta1s = np.linspace(-2,2, 100)
COST = np.empty(shape=(100,100))
# Meshgrid for paramaters
T0S, T1S = np.meshgrid(theta0s, theta1s)
# for each parameter combination compute the cost
for i in range(100):
for j in range(100):
COST[i,j] = cost(T0S[0,i], T1S[j,0], pga.distance, pga.accuracy)

# make 3d plot
fig2 = plt.figure()
ax = fig2.gca(projection='3d')
ax.plot_surface(X=T0S,Y=T1S,Z=COST)
plt.show()``````
``````# 对 theta1 进行求导# 对 thet
def partial_cost_theta1(theta0, theta1, x, y):
# Hypothesis
h = theta0 + theta1*x
# Hypothesis minus observed times x
diff = (h - y) * x
# Average to compute partial derivative
partial = diff.sum() / (x.shape)
return partial

partial1 = partial_cost_theta1(0, 5, pga.distance, pga.accuracy)
print("partial1 =", partial1)

# 对theta0 进行求导
# Partial derivative of cost in terms of theta0
def partial_cost_theta0(theta0, theta1, x, y):
# Hypothesis
h = theta0 + theta1*x
# Difference between hypothesis and observation
diff = (h - y)
# Compute partial derivative
partial = diff.sum() / (x.shape)
return partial

partial0 = partial_cost_theta0(1, 1, pga.distance, pga.accuracy)
print("partial0 =", partial0)``````

partial1 = 5.5791338540719
partial0 = 1.0000000000000104

``````# x is our feature vector -- distance
# y is our target variable -- accuracy
# alpha is the learning rate
# theta0 is the intial theta0
# theta1 is the intial theta1
def gradient_descent(x, y, alpha=0.1, theta0=0, theta1=0):
max_epochs = 1000 # Maximum number of iterations 最大迭代次数
counter = 0       # Intialize a counter 当前第几次
c = cost(theta1, theta0, pga.distance, pga.accuracy)  ## Initial cost 当前代价函数
costs = [c]     # Lets store each update 每次损失值都记录下来
# Set a convergence threshold to find where the cost function in minimized
# When the difference between the previous cost and current cost
#        is less than this value we will say the parameters converged
# 设置一个收敛的阈值 (两次迭代目标函数值相差没有相差多少,就可以停止了)
convergence_thres = 0.000001
cprev = c + 10
theta0s = [theta0]
theta1s = [theta1]

# When the costs converge or we hit a large number of iterations will we stop updating
# 两次间隔迭代目标函数值相差没有相差多少(说明可以停止了)
while (np.abs(cprev - c) > convergence_thres) and (counter < max_epochs):
cprev = c
# Alpha times the partial deriviative is our updated
# 先求导, 导数相当于步长
update0 = alpha * partial_cost_theta0(theta0, theta1, x, y)
update1 = alpha * partial_cost_theta1(theta0, theta1, x, y)

# Update theta0 and theta1 at the same time
# We want to compute the slopes at the same set of hypothesised parameters
#             so we update after finding the partial derivatives
# -= 梯度下降，+=梯度上升
theta0 -= update0
theta1 -= update1

# Store thetas
theta0s.append(theta0)
theta1s.append(theta1)

# Compute the new cost
# 当前迭代之后，参数发生更新
c = cost(theta0, theta1, pga.distance, pga.accuracy)

costs.append(c)
counter += 1   # Count

# 将当前的theta0, theta1, costs值都返回去
return {'theta0': theta0, 'theta1': theta1, "costs": costs}