[ Lecture Notes]Understanding Machine Learning

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

本帖隐藏的内容

http://www.cs.huji.ac.il/~shais/IML2014.html

本帖隐藏的内容

Shai Shalev-Shwartz.zip (9.94 MB)

2016-6-24 09:24:50 上传

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏3 回帖

关键词：Understand Learning machine earning Lecture

1. # -*- coding: utf-8 -*-
   
2. """
   
3. Simple demonstration of the Ellipsoid online learner on mnist
   
4. @author: shais
   
5. """
   
6. 
   
7. import numpy as np
   
8. import matplotlib.pyplot as plt
   
9. 
   
10. dot = np.dot;
    
11. sign = np.sign;
    
12. outer = np.outer;
    
13. zeros = np.zeros;
    
14. eye = np.eye;
    
15. sqrt = np.sqrt;
    
16. 
    
17. #%%
    
18. # read data
    
19. 
    
20. dataDir = "/Users/shais/data/mnist/mnist/";
    
21. 
    
22. X = np.loadtxt(dataDir + "train4vs7_data.txt.gz");
    
23. Y = np.loadtxt(dataDir + "train4vs7_labels.txt.gz");
    
24. d,n = X.shape;
    
25. 
    
26. #%%
    
27. 
    
28. # show some images
    
29. plt.figure(1);
    
30. for i in range (1,26):
    
31.     ax = plt.subplot(5,5,i);
    
32.     ax.axis('off');
    
33.     if Y[i]>0:
    
34.         ax.imshow(X[:,i].reshape(28,28),cmap="gray");
    
35.     else:
    
36.         ax.imshow(255-X[:,i].reshape(28,28),cmap="gray");
    
37. 
    
38. plt.draw();
    
39. 
    
40. #%%   
    
41. 
    
42. # Initial Ellipsoid learner
    
43. w = zeros((d,));
    
44. A = eye(d);
    
45. M = 0; # counts mistakes
    
46. 
    
47. #%%
    
48. 
    
49. # Loop Ellipsoid over data
    
50. eta = d*d/(d*d-1.0);
    
51. for t in range(0,n):
    
52.     yhat = sign(dot(w,X[:,t]));
    
53.     if Y[t] != yhat:
    
54.         M = M+1;
    
55.         Ax = dot(A , X[:,t]);
    
56.         xAx = dot(X[:,t] , Ax);
    
57.         w = w + Y[t]/((d+1)*sqrt(xAx)) * Ax;
    
58.         A = eta*( A - (2.0/((d+1.0)*xAx)) * outer(Ax,Ax) );
    
59. 
    
60. #%%
    
61. 
    
62. # show the mask learnt by ellipsoid
    
63. 
    
64. plt.figure(2);
    
65. ax1 = plt.subplot(1,2,1);
    
66. ax1.axis('off'); # no need for axis marks
    
67. ax2 = plt.subplot(1,2,2);
    
68. ax2.axis('off'); # no need for axis marks
    
69. ax1.imshow(w.reshape(28,28),cmap="gray");
    
70. tmp = 1/(1+np.exp(-10*w/w.max()));
    
71. ax2.imshow(tmp.reshape(28,28),cmap="gray");
    
72. plt.draw();
    
73. #%%

复制代码

Understanding Machine Learning

Lisrelchen 发表于 2016-6-24 09:18
**** 本内容被作者隐藏 ****
**** 本内容被作者隐藏 ****

谢谢分享

thanks for learning

kankan

谢谢分享

顶！！！！！

上课地方就是谢谢