1 import numpy as np
 2 import pylab as pl   #画图用
 3 from sklearn import svm
 4
 5 #随机生成两组二位数据
 6 np.random.seed(0)#使每次产生随机数不变
 7 X = np.r_[np.random.randn(20,2)-[2,2],np.random.randn(20,2)+[2,2]]#注意这里np.r_[],而不是np.r_（）我都打错了，会报错TypeError: 'RClass' object is not callable
 8 #np.r_是按列连接两个矩阵，就是把两矩阵上下相加，要求列数相等，np.c_是按行连接两个矩阵，就是把两矩阵左右相加，要求行数相等
 9
10 Y = [0] * 20+[1] * 20#Python原来可以这么简单的创建重复的列表呀
11
12 clf=svm.SVC(kernel='linear')
13 clf.fit(X,Y)
14
15 w=clf.coef_[0]
16 a=-w[0]/w[1]
17 xx=np.linspace(-5,5)#产生-5到5的线性连续值，间隔为1
18 yy=a*xx-(clf.intercept_[0])/w[1]  #clf.intercept_[0]是w3.即为公式a1*x1+a2*x2+w3中的w3。(clf.intercept_[0])/w[1]即为直线的截距
19
20 #得出支持向量的方程
21 b=clf.support_vectors_[0]
22 yy_down=a*xx+(b[1]-a*b[0])#(b[1]-a*b[0])就是简单的算截距
23 b=clf.support_vectors_[-1]
24 yy_up=a*xx+(b[1]-a*b[0])
25
26 print("w:",w) #打印出权重系数
27 print("a:",a) #打印出斜率
28 print("suport_vectors_:",clf.support_vectors_)#打印出支持向量
29 print("clf.coef_:",clf.coef_)                  #打印出权重系数，还是w
30
31
32 #这个就是画出来而已。很简单，也不太常用，都用matplotlib去了。不多说了
33 pl.plot(xx,yy,'k-')
34 pl.plot(xx,yy_down,'k--')
35 pl.plot(xx,yy_up,'k--')
36
37 pl.scatter(clf.support_vectors_[:,0],clf.support_vectors_[:,0],s=80,facecolors='none')
38 pl.scatter(X[:,0],X[:,1],c=Y,cmap=pl.cm.Paired)
39
40 pl.axis('tight')
41 pl.show()

 1 #from __future__ import print_function #__future__模块，把下一个新版本的特性导入到当前版本，于是我们就可以在当前版本中测试一些新版本的特性
 2                                         #我的Python版本是3.6.4.所以不需要这个
 3
 4 from time import time   #对程序运行时间计时用的
 5 import logging           #打印程序进展日志用的
 6 import matplotlib.pyplot as plt  #绘图用的
 7
 8 from sklearn.model_selection import train_test_split
 9 from sklearn.datasets import fetch_lfw_people
10 from sklearn.model_selection import GridSearchCV
11 from sklearn.metrics import classification_report
12 from sklearn.metrics import confusion_matrix
13 from sklearn.decomposition import PCA
14 from sklearn.svm import SVC
15
16 logging.basicConfig(level=logging.INFO,format='%(asctime)s %(message)s')
17 lfw_people=fetch_lfw_people(min_faces_per_person=70,resize=0.4)  #名人的人脸数据集、
18
19 n_samples,h,w=lfw_people.images.shape  #多少个实例，h,w高度，宽度值
20
21 X=lfw_people.data   #特征向量矩阵
22 n_feature=X.shape[1]#每个人有多少个特征值
23
24 Y=lfw_people.target
25 target_names=lfw_people.target_names
26 n_classes=target_names.shape[0]     #多少类
27 print("Total dataset size")
28 print("n_samples:",n_samples)
29 print("n_feature:",n_feature)
30 print("n_classes:",n_classes)
31
32 X_train,X_test,Y_train,Y_test=train_test_split(X,Y,test_size=0.25)  #选取0.25的测试集
33
34 #降维
35 n_components=150  #PCA算法中所要保留的主成分个数n，也即保留下来的特征个数n
36 print("Extracting the top %d eigenfaces from %d faces" % (n_components,X_train.shape[0]))
37 t0=time()
38 pca=PCA(svd_solver='randomized',n_components=n_components,whiten=True).fit(X_train)#训练一个pca模型
39
40 print("Train PCA in %0.3fs" % (time()-t0))
41
42 eigenfaces  = pca.components_.reshape((n_components,h,w))  #提取出来特征值之后的矩阵
43
44 print("Prijecting the input data on the eigenfaces orthonarmal basis")
45 t0=time()
46 X_train_pca = pca.transform(X_train)     #将训练集与测试集降维
47 X_test_pca = pca.transform(X_test)
48 print("Done PCA in %0.3fs" % (time()-t0))
49
50
51 #终于到SVM训练了
52 print("Fiting the classifier to the training set")
53 t0=time()
54 param_grid ={'C':[1e3,5e3,1e4,5e4,1e5],#C是对错误的惩罚
55              'gamma':[0.0001,0.0005,0.001,0.005,0.01,0.1],}#gamma核函数里多少个特征点会被使用}#对参数尝试不同的值
56 clf = GridSearchCV(SVC(kernel='rbf'),param_grid)
57 clf=clf.fit(X_train_pca,Y_train)
58 print("Done Fiting in %0.3fs" % (time()-t0))
59
60 print("Best estimotor found by grid search:")
61 print(clf.best_estimator_)
62
63 print("Predicting people's names on the test set")
64 t0=time()
65 Y_pred = clf.predict(X_test_pca)
66 print("done Predicting in %0.3fs" % (time()-t0))
67
68 print(classification_report(Y_test,Y_pred,target_names=target_names)) #生成一个小报告呀
69 print(confusion_matrix(Y_test,Y_pred,labels=range(n_classes)))#这个也是，生成的矩阵的意思是有多少个被分为此类。
70
71
72 #把分类完的图画出来12个。
73
74 #这个函数就是画图的
75 def plot_gallery(images,titles,h,w,n_row=3,n_col=4):
76     plt.figure(figsize=(1.8*n_col,2.4*n_row))
77     plt.subplots_adjust(bottom=0,left=.01,right=.99,top=.90,hspace=.35)
78     for i in range(n_row*n_col):
79         plt.subplot(n_row,n_col,i+1)
80         plt.imshow(images[i].reshape((h,w)),cmap=plt.cm.gray)
81         plt.title(titles[i],size=12)
82         plt.xticks(())
83         plt.yticks(())
84
85 #这个函数是生成一个固定格式的字符串的
86 def title(y_pred,y_test,target_names,i):
87     pred_name=target_names[y_pred[i]].rsplit(' ',1)[-1]
88     true_name = target_names[y_test[i]].rsplit(' ', 1)[-1]
89     return "predicted: %s\n true: %s" %(pred_name,true_name)
90
91 predicted_titles=[title(Y_pred,Y_test,target_names,i) for i in range(Y_pred.shape[0])] #这个for循环的用法很简介
92
93 plot_gallery(X_test,predicted_titles,h,w)
94
95 eigenfaces_titles=["eigenface %d " % i for i in range(eigenfaces.shape[0])]
96 plot_gallery(eigenfaces,eigenfaces_titles,h,w)
97
98 plt.show()

'''
看他的官方文档
def fetch_lfw_people(data_home=None, funneled=True, resize=0.5,min_faces_per_person=0, color=False,slice_=(slice(70, 195), slice(78, 172)),download_if_missing=True):"""Loader for the Labeled Faces in the Wild (LFW) people datasetThis dataset is a collection of JPEG pictures of famous peoplecollected on the internet, all details are available on theofficial website:http://vis-www.cs.umass.edu/lfw/Each picture is centered on a single face. Each pixel of each channel(color in RGB) is encoded by a float in range 0.0 - 1.0.The task is called Face Recognition (or Identification): given thepicture of a face, find the name of the person given a training set(gallery).
The original images are 250 x 250 pixels, but the default slice and resizearguments reduce them to 62 x 74.Parameters----------data_home : optional, default: NoneSpecify another download and cache folder for the datasets. By defaultall scikit learn data is stored in '~/scikit_learn_data' subfolders.funneled : boolean, optional, default: TrueDownload and use the funneled variant of the dataset.resize : float, optional, default 0.5Ratio used to resize the each face picture.min_faces_per_person : int, optional, default NoneThe extracted dataset will only retain pictures of people that have atleast `min_faces_per_person` different pictures.color : boolean, optional, default FalseKeep the 3 RGB channels instead of averaging them to a singlegray level channel. If color is True the shape of the data hasone more dimension than the shape with color = False.slice_ : optionalProvide a custom 2D slice (height, width) to extract the'interesting' part of the jpeg files and avoid use statisticalcorrelation from the backgrounddownload_if_missing : optional, True by defaultIf False, raise a IOError if the data is not locally availableinstead of trying to download the data from the source site.Returns-------dataset : dict-like object with the following attributes:dataset.data : numpy array of shape (13233, 2914)Each row corresponds to a ravelled face image of original size 62 x 47pixels. Changing the ``slice_`` or resize parameters will change theshape of the output.dataset.images : numpy array of shape (13233, 62, 47)Each row is a face image corresponding to one of the 5749 people inthe dataset. Changing the ``slice_`` or resize parameters will changethe shape of the output.dataset.target : numpy array of shape (13233,)Labels associated to each face image. Those labels range from 0-5748and correspond to the person IDs.dataset.DESCR : stringDescription of the Labeled Faces in the Wild (LFW) dataset意思就是min_faces_per_person=定义了每个人取多少个不同图片,resize=重新定义了图片的大小
'''

'''
n_components:PCA算法中所要保留的主成分个数n，也即保留下来的特征个数n
类型：int 或者 string，缺省时默认为None，所有成分被保留。赋值为int，比如n_components=1，将把原始数据降到一个维度。赋值为string，比如n_components='mle'，将自动选取特征个数n，使得满足所要求的方差百分比。copy:表示是否在运行算法时，将原始训练数据复制一份。若为True，则运行PCA算法后，原始训练数据的值不会有任何改变，因为是在原始数据的副本上进行运算；若为False，则运行PCA算法后，原始训练数据的值会改，因为是在原始数据上进行降维计算。
类型：bool，True或者False，缺省时默认为True。whiten:白化，使得每个特征具有相同的方差。由于图像中相邻像素之间具有很强的相关性，所以用于训练时输入是冗余的。白化的目的就是降低输入的冗余性；更正式的说，我们希望通过白化过程使得学习算法的输入具有如下性质：(i)特征之间相关性较低；(ii)所有特征具有相同的方差。类型：bool，缺省时默认为FalsePCA对象的属性
components_ ：返回具有最大方差的成分。
explained_variance_ratio_：返回 所保留的n个成分各自的方差百分比。
n_components_：返回所保留的成分个数n。
mean_：
noise_variance_：'''