【ML-SVM案例学习】002梯度下降之求解最优解

---

前言

【ML-SVM案例】会有十种SVM案例，供大家用来学习。本文只是实现梯度下降，求解最优解。后面一章将会实现003梯度下降：拉格乘子法。

---

提示：以下是本篇文章正文内容，下面案例可供参考

一、代码程序

1.引入库

代码如下（示例）：

import numpy as np
import matplotlib as mpl
import matplotlib.pyplot as plt
import math
from mpl_toolkits.mplot3d import Axes3D
1
2
3
4
5

2.一维原始图像与导函数

代码如下（示例）：

# 解决中文显示问题
mpl.rcParams['font.sans-serif'] = [u'SimHei']
mpl.rcParams['axes.unicode_minus'] = False

"""一维原始图像"""
def f1(x):
    return 0.5 * (x - 0.25) ** 2
# 导函数
def h1(x):
    return 0.5 * 2 * (x - 0.25)
1
2
3
4
5
6
7
8
9
10

3.使用梯度下降法求解

GD_X = []
GD_Y = []
x = 4
alpha = 0.5
f_change = f1(x)
f_current = f_change
GD_X.append(x)
GD_Y.append(f_current)
iter_num = 0
while f_change > 1e-10 and iter_num < 100:
    iter_num += 1
    x = x - alpha * h1(x)
    tmp = f1(x)
    f_change = np.abs(f_current - tmp)
    f_current  = tmp
    GD_X.append(x)
    GD_Y.append(f_current)
print(u"最终结果为:(%.5f, %.5f)" % (x, f_current))
print(u"迭代过程中X的取值，迭代次数:%d" % iter_num)
print(GD_X)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

4.构建数据与画图

X = np.arange(-4, 4.5, 0.05)
Y = np.array(list(map(lambda t: f1(t), X)))

plt.figure(facecolor='w')
plt.plot(X, Y, 'r-', linewidth=2)
plt.plot(GD_X, GD_Y, 'ko--', linewidth=2)
plt.title(u'函数$y=0.5 * (θ - 0.25)^2$; \n学习率:%.3f; 最终解:(%.3f, %.3f);迭代次数:%d' % (alpha, x, f_current, iter_num))
plt.show()
1
2
3
4
5
6
7
8

5.二维原始图像与导函数

"""二维原始图像"""
def f2(x, y):
    return 0.6 * (x + y) ** 2 - x * y
# 导函数
def hx2(x, y):
    return 0.6 * 2 * (x + y) - y
def hy2(x, y):
    return 0.6 * 2 * (x + y) - x
1
2
3
4
5
6
7
8

6. 使用梯度下降法求解

GD_X1 = []
GD_X2 = []
GD_Y = []
x1 = 4
x2 = 4
alpha = 0.5
f_change = f2(x1, x2)
f_current = f_change
GD_X1.append(x1)
GD_X2.append(x2)
GD_Y.append(f_current)
iter_num = 0
while f_change > 1e-10 and iter_num < 100:
    iter_num += 1
    prex1 = x1
    prex2 = x2
    x1 = x1 - alpha * hx2(prex1, prex2)
    x2 = x2 - alpha * hy2(prex1, prex2)
    
    tmp = f2(x1, x2)
    f_change = np.abs(f_current - tmp)
    f_current  = tmp
    GD_X1.append(x1)
    GD_X2.append(x2)
    GD_Y.append(f_current)
print(u"最终结果为:(%.5f, %.5f, %.5f)" % (x1, x2, f_current))
print(u"迭代过程中X的取值，迭代次数:%d" % iter_num)
print(GD_X1)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

7.构建数据与绘图

# 构建数据
X1 = np.arange(-4, 4.5, 0.2)
X2 = np.arange(-4, 4.5, 0.2)
X1, X2 = np.meshgrid(X1, X2)
Y = np.array(list(map(lambda t: f2(t[0], t[1]), zip(X1.flatten(), X2.flatten()))))
Y.shape = X1.shape


# 画图
fig = plt.figure(facecolor='w')
ax = Axes3D(fig)
ax.plot_surface(X1, X2, Y, rstride=1, cstride=1, cmap=plt.cm.jet)
ax.plot(GD_X1, GD_X2, GD_Y, 'ko--')

ax.set_title(u'函数$y=0.6 * (θ1 + θ2)^2 - θ1 * θ2$;\n学习率:%.3f; 最终解:(%.3f, %.3f, %.3f);迭代次数:%d' % (alpha, x1, x2, f_current, iter_num))
plt.show()
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

二、完整代码

import numpy as np
import matplotlib as mpl
import matplotlib.pyplot as plt
import math
from mpl_toolkits.mplot3d import Axes3D

# 解决中文显示问题
mpl.rcParams['font.sans-serif'] = [u'SimHei']
mpl.rcParams['axes.unicode_minus'] = False

"""一维原始图像"""
def f1(x):
    return 0.5 * (x - 0.25) ** 2
# 导函数
def h1(x):
    return 0.5 * 2 * (x - 0.25)

# 使用梯度下降法求解
GD_X = []
GD_Y = []
x = 4
alpha = 0.5
f_change = f1(x)
f_current = f_change
GD_X.append(x)
GD_Y.append(f_current)
iter_num = 0
while f_change > 1e-10 and iter_num < 100:
    iter_num += 1
    x = x - alpha * h1(x)
    tmp = f1(x)
    f_change = np.abs(f_current - tmp)
    f_current  = tmp
    GD_X.append(x)
    GD_Y.append(f_current)
print(u"最终结果为:(%.5f, %.5f)" % (x, f_current))
print(u"迭代过程中X的取值，迭代次数:%d" % iter_num)
print(GD_X)


# 构建数据
X = np.arange(-4, 4.5, 0.05)
Y = np.array(list(map(lambda t: f1(t), X)))

# 画图
plt.figure(facecolor='w')
plt.plot(X, Y, 'r-', linewidth=2)
plt.plot(GD_X, GD_Y, 'ko--', linewidth=2)
plt.title(u'函数$y=0.5 * (θ - 0.25)^2$; \n学习率:%.3f; 最终解:(%.3f, %.3f);迭代次数:%d' % (alpha, x, f_current, iter_num))
plt.show()


"""二维原始图像"""
def f2(x, y):
    return 0.6 * (x + y) ** 2 - x * y
# 导函数
def hx2(x, y):
    return 0.6 * 2 * (x + y) - y
def hy2(x, y):
    return 0.6 * 2 * (x + y) - x

# 使用梯度下降法求解
GD_X1 = []
GD_X2 = []
GD_Y = []
x1 = 4
x2 = 4
alpha = 0.5
f_change = f2(x1, x2)
f_current = f_change
GD_X1.append(x1)
GD_X2.append(x2)
GD_Y.append(f_current)
iter_num = 0
while f_change > 1e-10 and iter_num < 100:
    iter_num += 1
    prex1 = x1
    prex2 = x2
    x1 = x1 - alpha * hx2(prex1, prex2)
    x2 = x2 - alpha * hy2(prex1, prex2)
    
    tmp = f2(x1, x2)
    f_change = np.abs(f_current - tmp)
    f_current  = tmp
    GD_X1.append(x1)
    GD_X2.append(x2)
    GD_Y.append(f_current)
print(u"最终结果为:(%.5f, %.5f, %.5f)" % (x1, x2, f_current))
print(u"迭代过程中X的取值，迭代次数:%d" % iter_num)
print(GD_X1)


# 构建数据
X1 = np.arange(-4, 4.5, 0.2)
X2 = np.arange(-4, 4.5, 0.2)
X1, X2 = np.meshgrid(X1, X2)
Y = np.array(list(map(lambda t: f2(t[0], t[1]), zip(X1.flatten(), X2.flatten()))))
Y.shape = X1.shape


# 画图
fig = plt.figure(facecolor='w')
ax = Axes3D(fig)
ax.plot_surface(X1, X2, Y, rstride=1, cstride=1, cmap=plt.cm.jet)
ax.plot(GD_X1, GD_X2, GD_Y, 'ko--')

ax.set_title(u'函数$y=0.6 * (θ1 + θ2)^2 - θ1 * θ2$;\n学习率:%.3f; 最终解:(%.3f, %.3f, %.3f);迭代次数:%d' % (alpha, x1, x2, f_current, iter_num))
plt.show()


"""二维原始图像"""
def f2(x, y):
    return 0.15 * (x + 0.5) ** 2 + 0.25 * (y  - 0.25) ** 2 + 0.35 * (1.5 * x - 0.2 * y + 0.35 ) ** 2  
## 偏函数
def hx2(x, y):
    return 0.15 * 2 * (x + 0.5) + 0.25 * 2 * (1.5 * x - 0.2 * y + 0.35 ) * 1.5
def hy2(x, y):
    return 0.25 * 2 * (y  - 0.25) - 0.25 * 2 * (1.5 * x - 0.2 * y + 0.35 ) * 0.2

# 使用梯度下降法求解
GD_X1 = []
GD_X2 = []
GD_Y = []
x1 = 4
x2 = 4
alpha = 0.5
f_change = f2(x1, x2)
f_current = f_change
GD_X1.append(x1)
GD_X2.append(x2)
GD_Y.append(f_current)
iter_num = 0
while f_change > 1e-10 and iter_num < 100:
    iter_num += 1
    prex1 = x1
    prex2 = x2
    x1 = x1 - alpha * hx2(prex1, prex2)
    x2 = x2 - alpha * hy2(prex1, prex2)
    
    tmp = f2(x1, x2)
    f_change = np.abs(f_current - tmp)
    f_current  = tmp
    GD_X1.append(x1)
    GD_X2.append(x2)
    GD_Y.append(f_current)
print(u"最终结果为:(%.5f, %.5f, %.5f)" % (x1, x2, f_current))
print(u"迭代过程中X的取值，迭代次数:%d" % iter_num)
print(GD_X1)


# 构建数据
X1 = np.arange(-4, 4.5, 0.2)
X2 = np.arange(-4, 4.5, 0.2)
X1, X2 = np.meshgrid(X1, X2)
Y = np.array(list(map(lambda t: f2(t[0], t[1]), zip(X1.flatten(), X2.flatten()))))
Y.shape = X1.shape


# 画图
fig = plt.figure(facecolor='w')
ax = Axes3D(fig)
ax.plot_surface(X1, X2, Y, rstride=1, cstride=1, cmap=plt.cm.jet)
ax.plot(GD_X1, GD_X2, GD_Y, 'ko--')
ax.set_xlabel('x')
ax.set_ylabel('y')
ax.set_zlabel('z')

ax.set_title(u'函数;\n学习率:%.3f; 最终解:(%.3f, %.3f, %.3f);迭代次数:%d' % (alpha, x1, x2, f_current, iter_num))
plt.show()
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169

---

总结

001梯度下降：一维和二维图像
003梯度下降：拉格乘子法