版权声明:本文为博主原创文章,未经博主允许不得转载。 https://blog.csdn.net/mr_muli/article/details/84792169
- 机器学习之图半监督学习LabelSpreading
# -*- coding: utf-8 -*-
"""
Created on Tue Dec 4 13:32:30 2018
@author: muli
"""
import numpy as np
import matplotlib.pyplot as plt
from sklearn import metrics
from sklearn import datasets
from scipy.sparse.csgraph import connected_components
# 解决AttributeError: module 'scipy.sparse' has no attribute 'csgraph'问题
from sklearn.semi_supervised.label_propagation import LabelSpreading
def load_data():
'''
加载数据集
:return: 一个元组,依次为: 样本集合、样本标记集合、 未标记样本的下标集合
'''
digits = datasets.load_digits()
###### 混洗样本 ########
rng = np.random.RandomState(0)
indices = np.arange(len(digits.data)) # 样本下标集合
rng.shuffle(indices) # 混洗样本下标集合
X = digits.data[indices]
y = digits.target[indices]
###### 生成未标记样本的下标集合 ####
n_labeled_points = int(len(y)/10) # 只有 10% 的样本有标记
unlabeled_indices = np.arange(len(y))[n_labeled_points:] # 后面 90% 的样本未标记
return X,y,unlabeled_indices
def test_LabelSpreading(*data):
'''
测试 LabelSpreading 的用法
:param data: 一个元组,依次为: 样本集合、样本标记集合、 未标记样本的下标集合
:return: None
'''
X,y,unlabeled_indices=data
y_train=np.copy(y) # 必须拷贝,后面要用到 y
y_train[unlabeled_indices]=-1 # 未标记样本的标记设定为 -1
clf=LabelSpreading(max_iter=100,kernel='rbf',gamma=0.1)
clf.fit(X,y_train)
### 获取预测准确率
predicted_labels = clf.transduction_[unlabeled_indices] # 预测标记
true_labels = y[unlabeled_indices] # 真实标记
print("Accuracy:%f"%metrics.accuracy_score(true_labels,predicted_labels))
# 或者 print("Accuracy:%f"%clf.score(X[unlabeled_indices],true_labels))
def test_LabelSpreading_rbf(*data):
'''
测试 LabelSpreading 的 rbf 核时,预测性能随 alpha 和 gamma 的变化
:param data: 一个元组,依次为: 样本集合、样本标记集合、 未标记样本的下标集合
:return: None
'''
X,y,unlabeled_indices=data
y_train=np.copy(y) # 必须拷贝,后面要用到 y
y_train[unlabeled_indices]=-1 # 未标记样本的标记设定为 -1
fig=plt.figure()
ax=fig.add_subplot(1,1,1)
alphas=np.linspace(0.01,1,num=10,endpoint=False)
gammas=np.logspace(-2,2,num=50)
colors=((1,0,0),(0,1,0),(0,0,1),(0.5,0.5,0),(0,0.5,0.5),(0.5,0,0.5),
(0.4,0.6,0),(0.6,0.4,0),(0,0.6,0.4),(0.5,0.3,0.2),) # 颜色集合,不同曲线用不同颜色
## 训练并绘图
for alpha,color in zip(alphas,colors):
scores=[]
for gamma in gammas:
clf=LabelSpreading(max_iter=100,gamma=gamma,alpha=alpha,kernel='rbf')
clf.fit(X,y_train)
scores.append(clf.score(X[unlabeled_indices],y[unlabeled_indices]))
ax.plot(gammas,scores,label=r"$\alpha=%s$"%alpha,color=color)
### 设置图形
ax.set_xlabel(r"$\gamma$")
ax.set_ylabel("score")
ax.set_xscale("log")
ax.legend(loc="best")
ax.set_title("LabelSpreading rbf kernel")
plt.show()
def test_LabelSpreading_knn(*data):
'''
测试 LabelSpreading 的 knn 核时,预测性能随 alpha 和 n_neighbors 的变化
:param data: 一个元组,依次为: 样本集合、样本标记集合、 未标记样本的下标集合
:return: None
'''
X,y,unlabeled_indices=data
y_train=np.copy(y) # 必须拷贝,后面要用到 y
y_train[unlabeled_indices]=-1 # 未标记样本的标记设定为 -1
fig=plt.figure()
ax=fig.add_subplot(1,1,1)
alphas=np.linspace(0.01,1,num=10,endpoint=False)
Ks=[1,2,3,4,5,8,10,15,20,25,30,35,40,50]
colors=((1,0,0),(0,1,0),(0,0,1),(0.5,0.5,0),(0,0.5,0.5),(0.5,0,0.5),
(0.4,0.6,0),(0.6,0.4,0),(0,0.6,0.4),(0.5,0.3,0.2),) # 颜色集合,不同曲线用不同颜色
## 训练并绘图
for alpha,color in zip(alphas,colors):
scores=[]
for K in Ks:
clf=LabelSpreading(kernel='knn',max_iter=100,n_neighbors=K,alpha=alpha)
clf.fit(X,y_train)
scores.append(clf.score(X[unlabeled_indices],y[unlabeled_indices]))
ax.plot(Ks,scores,label=r"$\alpha=%s$"%alpha,color=color)
### 设置图形
ax.set_xlabel(r"$k$")
ax.set_ylabel("score")
ax.legend(loc="best")
ax.set_title("LabelSpreading knn kernel")
plt.show()
if __name__=='__main__':
data=load_data() # 获取半监督分类数据集
# test_LabelSpreading(*data) # 调用 test_LabelSpreading
# test_LabelSpreading_rbf(*data)# 调用 test_LabelSpreading_rbf
test_LabelSpreading_knn(*data)# 调用 test_LabelSpreading_knn
- 如图: