文字内容
1. Machine Learning Basics
2. OUTLINE 1. Learning Problems 2. Theory of Generalization 3. Over-fitting & Validation 4. Parameter Estimation 5. Supervised and Unsupervised Learning Algorithms 6. Challenges Motivating Deep Learning
3. REFERENCE
5. 1. Learning Problem
6. 1.1 What is Machine Learning Learning Problem 1.2 Key Essence of Machine Learning 1.3 Learning Flow 1.4 Type of Learning
7. 1.1 What is ML A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P , if its performance at tasks in T , as measured by P , improves with experience E . —— Mitchell(1997) 1. Learning Problem
8. 1.1 What is ML 1. Learning Problem
9. 1.1 What is ML 1. Learning Problem
10. 1.2 Key Essence of Machine Learnin g 1. Learning Problem
11. 1.3 Components of Machine Learnin g 1. Learning Problem
12. 1.3 Components of Machine Learnin g 1. Learning Problem
13. 1.4 Type of Learning 1. Learning Problem
14. 1.4 Type of Learning 1. Learning Problem
15. 1.4 Type of Learning supervised Learning unsupervised Learning 1. Learning Problem
16. 1.4 Type of Learning 1. Learning Problem
17. 1.4 Type of Learning
18. 2. Theory of Generalizatio n
19. 2.1 No Free Lunch Theory of Generalization 2.2 VC Theory 2.3 Bias &Variance 2.4 Learning Curve
21. 2.1 No Free Lunch 2. Theory of Generalization
22. 2.1 No Free Lunch 2. Theory of Generalization
23. 2.2 VC Theory 2. Theory of Generalization
24. 2.2 VC Theory
25. 2.2 VC Theory
27. 2.2 VC Theory
28. 2.2 VC Theory
29. 2.2 VC Theory
30. 2.2 VC Theory
31. 2.2 VC Theory
32. 2.2 VC Theory
33. 2.2 VC Theory
34. 2.2 VC Theory
35. 2.2 VC Theory K
36. 2.2 VC Theory
37. 2.2 VC Theory
38. 2.2 VC Theory
39. 2.2 VC Theory Looseness of VC Bound
40. 2.2 VC Theory
41. 2.2 VC Theory
42. 2.2 VC Theory
43. 2.3 Bias & Variance
45. 2.3 Bias & Variance
46. 2.3 Bias & Variance
47. 2.3 Learning Curve 2. Theory of Generalization
48. 3. Over-fitting & Validation
49. Over-fitting & Validation 3.1 What is over-fitting 3.2 The reason of over-fitting 3.3 How to avoid over-fitting
50. 3.1 What is over-fitting 3. Over-fitting & Validation
51. 3.1 What is over-fitting
52. 3.1 What is over-fitting
53. 3.1 What is over-fitting 3. Over-fitting & Validation
54. 3.1 What is over-fitting
55. 3.1 What is over-fitting
56. 3.2 The reason of over-fitting
58. 3.2 The reason of over-fitting
59. 3.2 The reason of over-fitting The error incurred by an oracle making predictions from the true distribution p(x, y) is called the Bayes error
60. 3.2 The reason of over-fitting
61. 3.2 The reason of over-fitting
64. 3.2 The reason of over-fitting The error incurred by an oracle making predictions from the true distribution p(x, y) is called the Bayes error
67. 3.2 The reason of over-fitting 1)Use excessive d_vc 2)Noise 3)Limited data size N
68. 3.3 How to avoid over-fitting 1)Use excessive d_vc 1) Start from simple model 2)Noise 2) data cleaning/pruning 3)Limited data size3) N data hinting 4)Regularization 5)Validation 3. Over-fitting & Validation
69. 3.3 How to avoid overfitting[Regularization] 3. Over-fitting & Validation
70. 3.3 How to avoid overfitting[Regularization]
71. 3.3 How to avoid overfitting[Regularization]
72. 3.3 How to avoid over-fitting [Regula rization]
73. 3.3 How to avoid overfitting[Regularization]
74. 3.3 How to avoid overfitting[Regularization]
76. 3.3 How to avoid overfitting[Regularization]
77. 3.3 How to avoid overfitting[Regularization]
78. 3.3 How to avoid overfitting[Regularization] ��� (N) ���� ���� (k) Large k ������ Small k (N-k) � �= 5
79. 3.3 How to avoid overfitting[Regularization]
80. 3.3 How to avoid overfitting[Regularization]
81. 3.3 How to avoid overfitting[Regularization]
82. 3.3 How to avoid overfitting[Regularization] Leave one out : K=1 Cross validation
83. 3.3 How to avoid overfitting[Regularization]
84. 3.3 How to avoid overfitting[Regularization]