-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathsvm_train.py
96 lines (89 loc) · 3.76 KB
/
svm_train.py
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
#encoding:utf-8
'''
使用线性SVM与高斯核的SVM分类西瓜数据集3.0α
'''
import os
import sys
import itertools
basedir = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
sys.path.append(basedir)
from lib.libsvm import svmutil
from dataSet.watermelon_3alpha import watermelon_counterexample_x, watermelon_posiexam_x
import matplotlib.pyplot as plt
import numpy as np
from scipy.optimize import fsolve
from functools import partial, reduce
DATA_PATH = basedir + r"/dataSet/watermelon3.0alpha.libsvm"
def plot_init():
'''
初始化matplot
'''
(ax1, ax2) = plt.subplots(2, 1)[1]
ax1.set_title(u'Liner SVM(Top),Gaussian SVM(Bottom)')
ax1.set_ylabel(u"Sugar Rate")
ax1.set_xlim(0, 1)
ax1.set_ylim(0, 0.7)
pose_points, counter_points = zip(*watermelon_posiexam_x), zip(*watermelon_counterexample_x)
ax1.plot(next(pose_points), next(pose_points), 'go')
ax1.plot(next(counter_points), next(counter_points), 'bo')#样本点
ax2.set_xlabel(u"Density")
ax2.set_ylabel(u"Sugar Rate")
ax2.set_xlim(0, 1)
ax2.set_ylim(0, 0.7)
pose_points, counter_points = zip(*watermelon_posiexam_x), zip(*watermelon_counterexample_x)
ax2.plot(next(pose_points), next(pose_points), 'go')
ax2.plot(next(counter_points), next(counter_points), 'bo')#样本点
return ax1, ax2
def decision_gaussian(svs, coef, gamma, bias):
'''
根据高斯函数的参数计算结果
输入 支持向量， 权值， gamma， 偏差
返回一个决策函数
'''
def func(feature_vector):
unitfunc = lambda coefunit, sv: coefunit[0] * np.exp(-gamma* np.sum(np.square(np.array(sv) - np.array(feature_vector)), axis=1))
return sum(map(unitfunc, coef, svs)) + np.array(bias)
return func
def main():
'''
主函数
'''
# 载入数据集
assert os.path.exists(DATA_PATH), r"数据集不存在"
labels, feature_spaces = svmutil.svm_read_problem(DATA_PATH)
#1 线性核SVM
liner_model = svmutil.svm_train(labels, feature_spaces, '-h 0 -t 0 -c 100 -q')
wm_sv, wm_coef = liner_model.get_SV(), liner_model.get_sv_coef()
liner_omega = [0, 0]
for sv, coef in itertools.zip_longest(wm_sv, wm_coef):
liner_omega[0] += sv[1] * coef[0]
liner_omega[1] += sv[2] * coef[0]
liner_bias = liner_model.get_sv_rho()
ax1, ax2 = plot_init()
ax1.plot([-liner_bias/liner_omega[0], (liner_omega[1]+liner_bias)/(-liner_omega[0])], [0, 1])
#2 高斯核SVM
## 1 求高斯核参数
gaussian_model = svmutil.svm_train(labels, feature_spaces, '-h 0 -t 2 -c 1000 -q')
wm_sv, wm_coef = gaussian_model.get_SV(), gaussian_model.get_sv_coef()
gaussian_svs = [[feature_spaces[1], feature_spaces[2]] for feature_spaces in wm_sv]
gaussian_bias = gaussian_model.get_sv_rho()
gaussian_gamma = 0.5 #查看libsvm中，默认的γ为类别的倒数
## 2 求解决策函数边界（即结果为0的点）
densitys = np.arange(0, 1, 0.01)
decison_func = decision_gaussian(svs=gaussian_svs, coef=wm_coef, gamma=gaussian_gamma, bias=gaussian_bias)
func = lambda x0: fsolve(lambda y0:decison_func( np.array([[x0, y0[0]], [x0, y0[1]]])), [0, 1])
sugar_rates = list(map(func, densitys))
pairs = map(lambda item:list(itertools.product([item[0]], item[1])), zip(densitys, sugar_rates))
pairs = list(reduce(lambda density,sugar_rate:density+sugar_rate, pairs))
'''
pairs格式[(x1,x2),...]
'''
results = np.abs(decison_func(pairs))
filtered_pairs = [pairs[i] for i in range(len(pairs)) if np.abs(results[i]) < 1e-3 ]
'''
由于scipy在求根时，无解的时候也会返回结果，故采取措施滤掉无效的结果
'''
ax2.plot(*list(zip(*filtered_pairs)), '.r')
plt.show()
if __name__ == "__main__":
main()