rbf函数python rbf函数映射计算

高斯核函数RBF

5-11、高斯核函数RBF

创新互联建站是创新、创意、研发型一体的综合型网站建设公司,自成立以来公司不断探索创新,始终坚持为客户提供满意周到的服务,在本地打下了良好的口碑,在过去的10年时间我们累计服务了上千家以及全国政企客户,如酒楼设计等企业单位,完善的项目管理流程,严格把控项目进度与质量监控加上过硬的技术实力获得客户的一致表扬。

import numpy as np

import matplotlib.pyplot as plt

from sklearn import datasets

from matplotlib.colors import ListedColormap

from sklearn.preprocessing import StandardScaler

from sklearn.svm import SVC

from sklearn.pipeline import Pipeline

from sklearn.model_selection import train_test_split

plt.rcParams['font.sans-serif'] = ['SimHei']

plt.rcParams['axes.unicode_minus'] = False

x, y = datasets.make_moons(n_samples=1000, noise=0.25, random_state=2020)  # 生成1000个数据样本

plt.figure()

plt.scatter(x[y == 0, 0], x[y == 0, 1], color="r")

plt.scatter(x[y == 1, 0], x[y == 1, 1], color="g")

plt.title('散点图')

plt.show()

x_train, x_test, y_train, y_test = train_test_split(x, y, random_state=2020)

# 绘制边界曲线

def plot_decision_boundary(model, axis):

x0, x1 = np.meshgrid(

    np.linspace(axis[0], axis[1], int((axis[1] - axis[0]) * 100)).reshape(-1, 1),

    np.linspace(axis[2], axis[3], int((axis[3] - axis[2]) * 100)).reshape(-1, 1)

)

x_new = np.c_[x0.ravel(), x1.ravel()]

y_pre = model.predict(x_new)

zz = y_pre.reshape(x0.shape)

# 设置颜色

cus = ListedColormap(["#BA55D3", "#FF69B4", "#FFE4C4"])

plt.contourf(x0, x1, zz, cmap=cus)

def RBFkernelSVC(gamma):#高斯核函数RBF

return Pipeline([

    ("std", StandardScaler()),

    ("svc", SVC(kernel="rbf", gamma=gamma))

])

sv = RBFkernelSVC(gamma=1)

sv.fit(x_train, y_train)

plot_decision_boundary(sv, axis=([-1.8, 2.5, -1.4, 1.8]))

plt.scatter(x[y == 0, 0], x[y == 0, 1], color="r")

plt.scatter(x[y == 1, 0], x[y == 1, 1], color="g")

plt.title('高斯核函数RBF')

plt.show()

# 打印出分数

print(sv.score(x_test, y_test))

d = datasets.load_iris()

x = d.data

y = d.target

x_train, x_test, y_train, y_test = train_test_split(x, y, random_state=2020)

sv = RBFkernelSVC(gamma=10)

sv.fit(x_train, y_train)

# 打印出分数

print(sv.score(x_test, y_test))

python rbf表示什么分布

径向基(RBF)神经网络python实现

1 from numpy import array, append, vstack, transpose, reshape, \

2                   dot, true_divide, mean, exp, sqrt, log, \

3                   loadtxt, savetxt, zeros, frombuffer

4 from numpy.linalg import norm, lstsq

5 from multiprocessing import Process, Array

6 from random import sample

7 from time import time

8 from sys import stdout

9 from ctypes import c_double

10 from h5py import File

11

12

13 def metrics(a, b):

14     return norm(a - b)

15

16

17 def gaussian (x, mu, sigma):

18     return exp(- metrics(mu, x)**2 / (2 * sigma**2))

21 def multiQuadric (x, mu, sigma):

22     return pow(metrics(mu,x)**2 + sigma**2, 0.5)

23

24

25 def invMultiQuadric (x, mu, sigma):

26     return pow(metrics(mu,x)**2 + sigma**2, -0.5)

27

28

29 def plateSpine (x,mu):

30     r = metrics(mu,x)

31     return (r**2) * log(r)

32

33

34 class Rbf:

35     def __init__(self, prefix = 'rbf', workers = 4, extra_neurons = 0, from_files = None):

36         self.prefix = prefix

37         self.workers = workers

38         self.extra_neurons = extra_neurons

39

40         # Import partial model

41         if from_files is not None:          

42             w_handle = self.w_handle = File(from_files['w'], 'r')

43             mu_handle = self.mu_handle = File(from_files['mu'], 'r')

44             sigma_handle = self.sigma_handle = File(from_files['sigma'], 'r')

45          

46             self.w = w_handle['w']

47             self.mu = mu_handle['mu']

48             self.sigmas = sigma_handle['sigmas']

49          

50             self.neurons = self.sigmas.shape[0]

51

52     def _calculate_error(self, y):

53         self.error = mean(abs(self.os - y))

54         self.relative_error = true_divide(self.error, mean(y))

55

56     def _generate_mu(self, x):

57         n = self.n

58         extra_neurons = self.extra_neurons

59

60         # TODO: Make reusable

61         mu_clusters = loadtxt('clusters100.txt', delimiter='\t')

62

63         mu_indices = sample(range(n), extra_neurons)

64         mu_new = x[mu_indices, :]

65         mu = vstack((mu_clusters, mu_new))

66

67         return mu

68

69     def _calculate_sigmas(self):

70         neurons = self.neurons

71         mu = self.mu

72

73         sigmas = zeros((neurons, ))

74         for i in xrange(neurons):

75             dists = [0 for _ in xrange(neurons)]

76             for j in xrange(neurons):

77                 if i != j:

78                     dists[j] = metrics(mu[i], mu[j])

79             sigmas[i] = mean(dists)* 2

80                       # max(dists) / sqrt(neurons * 2))

81         return sigmas

82

83     def _calculate_phi(self, x):

84         C = self.workers

85         neurons = self.neurons

86         mu = self.mu

87         sigmas = self.sigmas

88         phi = self.phi = None

89         n = self.n

90

91

92         def heavy_lifting(c, phi):

93             s = jobs[c][1] - jobs[c][0]

94             for k, i in enumerate(xrange(jobs[c][0], jobs[c][1])):

95                 for j in xrange(neurons):

96                     # phi[i, j] = metrics(x[i,:], mu[j])**3)

97                     # phi[i, j] = plateSpine(x[i,:], mu[j]))

98                     # phi[i, j] = invMultiQuadric(x[i,:], mu[j], sigmas[j]))

99                     phi[i, j] = multiQuadric(x[i,:], mu[j], sigmas[j])

100                     # phi[i, j] = gaussian(x[i,:], mu[j], sigmas[j]))

101                 if k % 1000 == 0:

102                     percent = true_divide(k, s)*100

103                     print(c, ': {:2.2f}%'.format(percent))

104             print(c, ': Done')

105      

106         # distributing the work between 4 workers

107         shared_array = Array(c_double, n * neurons)

108         phi = frombuffer(shared_array.get_obj())

109         phi = phi.reshape((n, neurons))

110

111         jobs = []

112         workers = []

113

114         p = n / C

115         m = n % C

116         for c in range(C):

117             jobs.append((c*p, (c+1)*p + (m if c == C-1 else 0)))

118             worker = Process(target = heavy_lifting, args = (c, phi))

119             workers.append(worker)

120             worker.start()

121

122         for worker in workers:

123             worker.join()

124

125         return phi

126

127     def _do_algebra(self, y):

128         phi = self.phi

129

130         w = lstsq(phi, y)[0]

131         os = dot(w, transpose(phi))

132         return w, os

133         # Saving to HDF5

134         os_h5 = os_handle.create_dataset('os', data = os)

135

136     def train(self, x, y):

137         self.n = x.shape[0]

138

139         ## Initialize HDF5 caches

140         prefix = self.prefix

141         postfix = str(self.n) + '-' + str(self.extra_neurons) + '.hdf5'

142         name_template = prefix + '-{}-' + postfix

143         phi_handle = self.phi_handle = File(name_template.format('phi'), 'w')

144         os_handle = self.w_handle = File(name_template.format('os'), 'w')

145         w_handle = self.w_handle = File(name_template.format('w'), 'w')

146         mu_handle = self.mu_handle = File(name_template.format('mu'), 'w')

147         sigma_handle = self.sigma_handle = File(name_template.format('sigma'), 'w')

148

149         ## Mu generation

150         mu = self.mu = self._generate_mu(x)

151         self.neurons = mu.shape[0]

152         print('({} neurons)'.format(self.neurons))

153         # Save to HDF5

154         mu_h5 = mu_handle.create_dataset('mu', data = mu)

155

156         ## Sigma calculation

157         print('Calculating Sigma...')

158         sigmas = self.sigmas = self._calculate_sigmas()

159         # Save to HDF5

160         sigmas_h5 = sigma_handle.create_dataset('sigmas', data = sigmas)

161         print('Done')

162

163         ## Phi calculation

164         print('Calculating Phi...')

165         phi = self.phi = self._calculate_phi(x)

166         print('Done')

167         # Saving to HDF5

168         print('Serializing...')

169         phi_h5 = phi_handle.create_dataset('phi', data = phi)

170         del phi

171         self.phi = phi_h5

172         print('Done')

173

174         ## Algebra

175         print('Doing final algebra...')

176         w, os = self.w, _ = self._do_algebra(y)

177         # Saving to HDF5

178         w_h5 = w_handle.create_dataset('w', data = w)

179         os_h5 = os_handle.create_dataset('os', data = os)

180

181         ## Calculate error

182         self._calculate_error(y)

183         print('Done')

184

185     def predict(self, test_data):

186         mu = self.mu = self.mu.value

187         sigmas = self.sigmas = self.sigmas.value

188         w = self.w = self.w.value

189

190         print('Calculating phi for test data...')

191         phi = self._calculate_phi(test_data)

192         os = dot(w, transpose(phi))

193         savetxt('iok3834.txt', os, delimiter='\n')

194         return os

195

196     @property

197     def summary(self):

198         return '\n'.join( \

199             ['-----------------',

200             'Training set size: {}'.format(self.n),

201             'Hidden layer size: {}'.format(self.neurons),

202             '-----------------',

203             'Absolute error   : {:02.2f}'.format(self.error),

204             'Relative error   : {:02.2f}%'.format(self.relative_error * 100)])

205

206

207 def predict(test_data):

208     mu = File('rbf-mu-212243-2400.hdf5', 'r')['mu'].value

209     sigmas = File('rbf-sigma-212243-2400.hdf5', 'r')['sigmas'].value

210     w = File('rbf-w-212243-2400.hdf5', 'r')['w'].value

211

212     n = test_data.shape[0]

213     neur = mu.shape[0]

214  

215     mu = transpose(mu)

216     mu.reshape((n, neur))

217

218     phi = zeros((n, neur))

219     for i in range(n):

220         for j in range(neur):

221             phi[i, j] = multiQuadric(test_data[i,:], mu[j], sigmas[j])

222

223     os = dot(w, transpose(phi))

224     savetxt('iok3834.txt', os, delimiter='\n')

225     return os

python3.5做分类时,混淆矩阵加在哪一步

preface:做着最近的任务,对数据处理,做些简单的提特征,用机器学习算法跑下程序得出结果,看看哪些特征的组合较好,这一系列流程必然要用到很多函数,故将自己常用函数记录上。应该说这些函数基本上都会用到,像是数据预处理,处理完了后特征提取、降维、训练预测、通过混淆矩阵看分类效果,得出报告。

1.输入

从数据集开始,提取特征转化为有标签的数据集,转为向量。拆分成训练集和测试集,这里不多讲,在上一篇博客中谈到用StratifiedKFold()函数即可。在训练集中有data和target开始。

2.处理

[python] view plain copy

def my_preprocessing(train_data):

from sklearn import preprocessing

X_normalized = preprocessing.normalize(train_data ,norm = "l2",axis=0)#使用l2范式,对特征列进行正则

return X_normalized

def my_feature_selection(data, target):

from sklearn.feature_selection import SelectKBest

from sklearn.feature_selection import chi2

data_new = SelectKBest(chi2, k= 50).fit_transform(data,target)

return data_new

def my_PCA(data):#data without target, just train data, withou train target.

from sklearn import decomposition

pca_sklearn = decomposition.PCA()

pca_sklearn.fit(data)

main_var = pca_sklearn.explained_variance_

print sum(main_var)*0.9

import matplotlib.pyplot as plt

n = 15

plt.plot(main_var[:n])

plt.show()

def clf_train(data,target):

from sklearn import svm

#from sklearn.linear_model import LogisticRegression

clf = svm.SVC(C=100,kernel="rbf",gamma=0.001)

clf.fit(data,target)

#clf_LR = LogisticRegression()

#clf_LR.fit(x_train, y_train)

#y_pred_LR = clf_LR.predict(x_test)

return clf

def my_confusion_matrix(y_true, y_pred):

from sklearn.metrics import confusion_matrix

labels = list(set(y_true))

conf_mat = confusion_matrix(y_true, y_pred, labels = labels)

print "confusion_matrix(left labels: y_true, up labels: y_pred):"

print "labels\t",

for i in range(len(labels)):

print labels[i],"\t",

print

for i in range(len(conf_mat)):

print i,"\t",

for j in range(len(conf_mat[i])):

print conf_mat[i][j],'\t',

print

print

def my_classification_report(y_true, y_pred):

from sklearn.metrics import classification_report

print "classification_report(left: labels):"

print classification_report(y_true, y_pred)

my_preprocess()函数:

主要使用sklearn的preprocessing函数中的normalize()函数,默认参数为l2范式,对特征列进行正则处理。即每一个样例,处理标签,每行的平方和为1.

my_feature_selection()函数:

使用sklearn的feature_selection函数中SelectKBest()函数和chi2()函数,若是用词袋提取了很多维的稀疏特征,有必要使用卡方选取前k个有效的特征。

my_PCA()函数:

主要用来观察前多少个特征是主要特征,并且画图。看看前多少个特征占据主要部分。

clf_train()函数:

可用多种机器学习算法,如SVM, LR, RF, GBDT等等很多,其中像SVM需要调参数的,有专门调试参数的函数如StratifiedKFold()(见前几篇博客)。以达到最优。

my_confusion_matrix()函数:

主要是针对预测出来的结果,和原来的结果对比,算出混淆矩阵,不必自己计算。其对每个类别的混淆矩阵都计算出来了,并且labels参数默认是排序了的。

my_classification_report()函数:

主要通过sklearn.metrics函数中的classification_report()函数,针对每个类别给出详细的准确率、召回率和F-值这三个参数和宏平均值,用来评价算法好坏。另外ROC曲线的话,需要是对二分类才可以。多类别似乎不行。

主要参考sklearn官网

利用RBF作为核函数

5-2、利用RBF作为核函数

import numpy as np

import matplotlib.pyplot as plt

from sklearn import svm, datasets

plt.rcParams['font.sans-serif'] = ['SimHei']

plt.rcParams['axes.unicode_minus'] = False

iris = datasets.load_iris()

# 为简单起见,选取前两个特征作为分类的输入特征,

# 以便在二维空间画出决策曲线

X = iris.data[:, :2]

y = iris.target

# 设置分类器SVC,核函数为rbf,gamma设置为自动调整

svc = svm.SVC(kernel="rbf", C=1, gamma="auto").fit(X, y)

# 绘图参数

x_min, x_max = X[:, 0].min() - 1, X[:, 0].max() + 1

y_min, y_max = X[:, 1].min() - 1, X[:, 1].max() + 1

h = (x_max / x_min) / 100

xx, yy = np.meshgrid(np.arange(x_min, x_max, h),

                np.arange(y_min, y_max, h))

plt.subplot(1, 1, 1)

# 利用已有分类器进行预测

Z = svc.predict(np.c_[xx.ravel(), yy.ravel()])

Z = Z.reshape(xx.shape)

# 绘制等高线并填充轮廓

plt.contourf(xx, yy, Z, cmap=plt.cm.Paired, alpha=0.8)

plt.scatter(X[:, 0], X[:, 1], c=y, cmap=plt.cm.Paired)

plt.xlabel('花萼长度')

plt.ylabel('花萼宽度')

# 限制x的取值范围,便于显示

plt.xlim(xx.min(), xx.max())

plt.title('利用RBF作为核函数')

plt.show()


文章标题:rbf函数python rbf函数映射计算
网页网址:http://ybzwz.com/article/doeohds.html