半监督学习官方案例

import numpy as np
import matplotlib.pyplot as plt
from scipy import stats
from sklearn import datasets
from sklearn.semi_supervised import LabelSpreading       // sklearn.semi_supervised中的半监督估计器

from sklearn.metrics import confusion_matrix, classification_report        //混淆矩阵、分类报告

digits = datasets.load_digits()   //导入手写体数字 数据集
rng = np.random.RandomState(2)     //生成随机数0-1796个数字
indices = np.arange(len(digits.data))  //切片
rng.shuffle(indices)              //打乱

X = digits.data[indices[:340]]      //取340个数据
y = digits.target[indices[:340]]
images = digits.images[indices[:340]]

n_total_samples = len(y)  //340
n_labeled_points = 40    //标注好的数据40条

indices = np.arange(n_total_samples)

unlabeled_set = indices[n_labeled_points:]  //未标注的数据
y_train = np.copy(y)
y_train[unlabeled_set] = -1    //将 y_train 中含有无标签的值设为 -1（进行清除）

//标签传播学习
lp_model = LabelSpreading(gamma=.25, max_iter=20)    //训练模型
lp_model.fit(X, y_train)
predicted_labels = lp_model.transduction_[unlabeled_set]   //预测标签
true_labels = y[unlabeled_set]   //真实标签

cm = confusion_matrix(true_labels, predicted_labels, labels=lp_model.classes_)

print("Label Spreading model: %d labeled & %d unlabeled points (%d total)" %
      (n_labeled_points, n_total_samples - n_labeled_points, n_total_samples))

print(classification_report(true_labels, predicted_labels))

print("Confusion matrix")
print(cm)

//计算每个转换分布的不确定值
pred_entropies = stats.distributions.entropy(lp_model.label_distributions_.T)

//选取前10个最不准的标签
uncertainty_index = np.argsort(pred_entropies)[-10:]

//绘图
f = plt.figure(figsize=(7, 5))   //设置图片大小
for index, image_index in enumerate(uncertainty_index):
    image = images[image_index]

    sub = f.add_subplot(2, 5, index + 1)   //设置子图
    sub.imshow(image, cmap=plt.cm.gray_r)
    plt.xticks([])
    plt.yticks([])
    sub.set_title(‘predict: %i\ntrue: %i‘ % (
        lp_model.transduction_[image_index], y[image_index]))