CS50 Machine Learning. Week 7

Transcription

1 CS5 Machine Learning Week 7

2 *pythonprogramming.net

3 Machine Learning?

4 Machine Learning? Search Engines Image Recognition Voice Recognition Natural Language Processing

5 inputs outputs

6 Image Recognition horse car

7 Natural Language Processing Nineteen Eighty-Four by George Orwell (1984) [...] BIG BROTHER IS WATCHING YOU, the caption said, while the dark eyes looked deep into Winston's own [...] Politics Propaganda Privacy

8 Whodunit! Image recognition horse car

9 Machine Learning algorithms inputs Training data outputs

10 Machine Learning algorithms Training data horse

11 Image Classification

12

13 Handwritten digit classification Training data

14 Nearest Neighbor Classifier Minimal distance? 6 66 Labeled training set Test point

15 Nearest Neighbor Classifier? 6 6 Minimal distance Labeled training set Test point 6 6

16 ?

17 Nearest Neighbor Classifier? Minimal distance Labeled training set Test point

18 ?

19

20 Flatland by Edwin Abbott Abbott (1884) *

21 Flatland, Edwin Abbott Abbott, 1984 Flatland: The story describes a two-dimensional world occupied by geometric figures. The narrator is a square named A Square who guides the readers through some of the implications of life in two dimensions. On New Year's Eve, A Square dreams about a visit to a one-dimensional world (Lineland) inhabited by "lustrous points", in which he attempts to convince the realm's monarch of a second dimension; but is unable to do so. Following this vision, A Square is himself visited by a three-dimensional sphere named A Sphere, which he cannot comprehend until he sees Spaceland (a tridimensional world) *

22 Ready to go beyond Lineland, Flatland, and Spaceland?

23 ?

24

25

26 dimensional space

27 Nearest Neighbor Classifier? 6 dist( Labeled training set 6, 6 6 ) Test point

28 ( dist( dist ,, ) ) = 31.98

29 ( dist( dist ,, ) ) = 45.97

30 The digits dataset Labeled training set

31 Python code (Supervised Learning)

32 np.sqrt(np.sum((x - y)**2))??? x y x (x = = - [1, 1] [3, 4] y = [-2, -3] y)**2 = [4, 9] np.sum((x - y)**2) = 13 np.sqrt(np.sum((x - y)**2)) = 3.6

33 Labeled Training subset Labeled training set Test point

34 Labeled Training set Testing set

35 Labeled Training set Testing set

36 With Nearest Neighbor Classifier 6 ~ 97% Correct

37 The CIFAR-1 dataset airplane automobile bird cat deer dog frog horse ship truck Labeled training set *

38 With Nearest Neighbor Classifier horse car ~ 3% Correct

39 Training set for category : Training set for category horse :

40 Challenges *

41 Features

42 Features (,,,)

43 Deep Learning *

44 Tensorflow Deep dream generator

45 The CIFAR-1 dataset airplane automobile bird cat deer dog frog horse ship truck Labeled training set *

46 With Deep Learning... horse car ~ 95% Correct

47 Is 95% enough?

48

49 MAY 216

50 Neither Autopilot nor the driver noticed the white side of the tractor trailer against a brightly lit sky, so the brake was not applied *

51 Challenges *

52 Text Clustering

53 Text clustering IMDB synopses for: - Robin Hood - The Matrix - The King's Speech - Aladdin - A Beautiful Mind - Finding Nemo CLUSTER 1: -??? A Beautiful Mind -??? The Matrix -??? The King's Speech CLUSTER 2: -??? Robin Hood -??? Aladdin -??? Finding Nemo k=2

54 k=2 Unlabeled data K-means

55 k=2 Unlabeled data K-means

56 Robin Hood Told with animals for it's cast, the story tells of Robin Hood (a fox) and Little John (a brown bear), who rob from the rich to give to the poor. [...]? Robin Hood

57 Unlabeled data k=2 A Beautiful Mind The Matrix Aladdin The King's Speech Robin Hood Finding Nemo K-means

58 Something simpler... a) I love CS5. Staff is awesome, awesome, awesome! b) I have a dog and a cat. c) Best of CS5? Staff. And cakes. Ok, CS5 staff. d) My dog keeps chasing my cat. Dogs! k=2 CLUSTER 1: a) c) CLUSTER 2: b) d)

59 k=2 b) I have a dog and a cat. d) My dog keeps chasing my cat. Dogs! a) I love CS5. Staff is awesome, awesome, awesome! c) Best of CS5? Staff. And cakes. Ok, CS5 staff. K-means

60 a) I love CS5. Staff is awesome, awesome, awesome!? a) I love CS5. Staff is awesome, awesome, awesome!

61 a) I love CS5. Staff is awesome, awesome, awesome! Bags of words b) I have a dog and a cat. c) Best of CS5? Staff. And cakes. Ok, CS5 staff. d) My dog keeps chasing my cat. Dogs! awesome best cakes cat chasing cs5 dog dogs keeps love ok staff a) b) 1 1 c) d)

62 a) I love CS5. Staff is awesome, awesome, awesome! b) I have a dog and a cat. c) Best of CS5? Staff. And cakes. Ok, CS5 staff. Frequency d) My dog keeps chasing my cat. Dogs! awesome best cakes cat chasing cs5 dog dogs keeps love ok staff a) 3/6 1/6 1/6 1 b) 1/2 1/2 c) 1/7 1/7 2/7 1/7 2/7 d) 1/5 1/5 1/5 1/5 1/5

63 a) I love CS5. Staff is awesome, awesome, awesome! a) I love CS5. Staff is awesome, awesome, awesome! (3/6,,,,, 1/6,,,, 1/6,, 1) 12 dimensional space

64 k=2 b) I have a dog and a cat. d) My dog keeps chasing my cat. Dogs! a) I love CS5. Staff is awesome, awesome, awesome! c) Best of CS5? Staff. And cakes. Ok, CS5 staff. K-means

65 Python code (Unsupervised Learning)

66 Recap

67 Handwritten digit classification 6

68 Text clustering IMDB synopses for: - Robin Hood - The Matrix - The King's Speech - Aladdin - A Beautiful Mind - Finding Nemo CLUSTER 1: - A Beautiful Mind - The Matrix - The King's Speech CLUSTER 2: - Robin Hood - Aladdin - Finding Nemo k=2

69 Machine Learning? Search Engines Image Recognition Voice Recognition Natural Language Processing

70 Machine Learning so much more # ## ### #### ##### ###### ####### # ## ### #### ##### ###### #######

71 Machine Learning so much more MARCH 216 Commentators were convinced [AlphaGo] had made mistakes, but as it racked up wins, they were forced to concede that perhaps the machine [...] was using strategies its human masters had simply overlooked. Lee Sedol *