Introduction to Machine Learning (UX Indonesia Meetup)

For UX ID Meetup

Transcript

1. Introduction to
   Machine Learning
   Galuh Sahid
   @galuhsahid | galuh.me
   UX Indonesia Meetup - June 23, 2020
   

   View Slide

2. • Data Scientist at Gojek
   • Google Developer Expert in Machine Learning
   • Co-host podcast Kartini Teknologi (kartiniteknologi.id)
   @galuhsahid
   Hi! I’m Galuh.
   

   View Slide

3. @galuhsahid
   Photo by Bram Van Oost from Unsplash
   

   View Slide

4. @galuhsahid
   Photo by Frank V. from Unsplash
   

   View Slide

5. @galuhsahid
   Photo by Bence Boros from Unsplash
   

   View Slide

6. @galuhsahid
   Photo by Nordwood from Unsplash
   

   View Slide

7. @galuhsahid
   Photo by Krsto Jevtic from Unsplash
   

   View Slide

8. It’s an exciting time to learn about
   machine learning!
   @galuhsahid
   

   View Slide

9. But… what is machine learning?
   @galuhsahid
   

   View Slide

10. A field of study that gives computers the ability to learn without being
    explicitly programmed.
    Arthur Samuel (1959)
    @galuhsahid
    

    View Slide

11. How is machine learning different
    from traditional programming?
    @galuhsahid
    

    View Slide

12. Traditional
    Programming
    Rules
    Data
    Answers
    @galuhsahid
    

    View Slide

13. if pixel[5][7] is black and pixel [5][6]
    is black and pixel [5][8] is black and …:
    if pixel[6][7] is black and pixel[6][7]
    is black and …:
    return “panda”
    …
    …
    …
    else:
    return “cat”
    Photo by Damian Patowski from Unsplash @galuhsahid
    

    View Slide

14. if pixel[5][7] is black and pixel [5][6]
    is black and pixel [5][7] is black and …:
    if pixel[6][7] is black and pixel[6][7]
    is black and …:
    return “panda”
    …
    …
    …
    else:
    return “not cat”
    Photo by Dušan Smetana from Unsplash @galuhsahid
    

    View Slide

15. Machine
    Learning
    Answers
    Data
    Rules
    

    View Slide

16. Answers
    Data
    Panda
    Cat
    Cat
    Photo by Max Baskalov and Zane Lee from Unsplash
    Panda
    @galuhsahid
    

    View Slide

17. ?
    Cat
    ?
    Photo by Cyrus Chew from Unsplash @galuhsahid
    

    View Slide

18. Machine learning? Artificial
    intelligence? Neural networks? Deep
    learning?
    @galuhsahid
    

    View Slide

19. Source @galuhsahid
    

    View Slide

20. Rule engine
    Knowledge
    Graphs
    Source @galuhsahid
    

    View Slide

21. Regression
    Decision Tree
    Random Forest
    Source @galuhsahid
    

    View Slide

22. Ingredients of machine learning
    @galuhsahid
    

    View Slide

23. Ingredient #1
    Data
    @galuhsahid
    

    View Slide

24. Tabular
    @galuhsahid
    

    View Slide

25. Text
    @galuhsahid
    

    View Slide

26. Sound
    @galuhsahid
    

    View Slide

27. Image
    @galuhsahid
    

    View Slide

28. Ingredient #2
    Features
    @galuhsahid
    

    View Slide

29. Examples of features for house price
    prediction
    Features
    @galuhsahid
    

    View Slide

30. Examples of features for house price
    prediction
    Features
    We want to predict this…
    @galuhsahid
    

    View Slide

31. Examples of features for house price
    prediction
    Features
    We want to predict this…
    …using these features
    @galuhsahid
    

    View Slide

32. What are the features for an image
    classification task?
    Features
    @galuhsahid
    

    View Slide

33. What are the features for an image
    classification task?
    Features
    @galuhsahid
    Source
    

    View Slide

34. Ingredient #3
    Model
    @galuhsahid
    

    View Slide

35. What is a “model”?
    Model
    - A model maps examples to predicted labels
    - It is defined by weights that are learned during the training process
    - Once trained, you can use it to make predictions about data that it has
    never seen before
    @galuhsahid
    

    View Slide

36. The training process
    Model
    - Iteration 1: 2*number of floors + 3*area size = predicted house price
    Model
    Data Predictions
    House #1:
    predicted: 200 million
    actual: 500 million
    difference: 300 million
    @galuhsahid
    

    View Slide

37. The training process
    Model
    - Iteration 1: 2*number of floors + 3*area size = predicted house price
    Model
    Data Predictions
    House #1:
    predicted: 400 million
    actual: 500 million
    difference: 100 million
    - Iteration 2: 4*number of floors + 6*area size = predicted house price
    @galuhsahid
    

    View Slide

38. The training process
    Model
    - Iteration 1: 2*number of floors + 3*area size = predicted house price
    Model
    Data Predictions
    House #1:
    predicted: 400 million
    actual: 500 million
    difference: 100 million
    - Iteration 2: 4*number of floors + 6*area size = predicted house price
    Our model does not get smart right
    away - it needs to be “trained”
    @galuhsahid
    

    View Slide

39. Types of machine learning problems
    & applications
    @galuhsahid
    

    View Slide

40. Type #1
    Binary classification
    Classifies input into one of two categories
    @galuhsahid
    

    View Slide

41. Spam or not spam
    Binary Classification
    @galuhsahid
    

    View Slide

42. Type #2
    Multi-class classification
    Classifies input into one of more than two categories
    @galuhsahid
    

    View Slide

43. Language prediction
    Multi-class Classification
    @galuhsahid
    

    View Slide

44. Type #3
    Regression
    Predicts a value on a continuous scale
    @galuhsahid
    

    View Slide

45. Visit duration & wait time estimates
    Regression
    @galuhsahid
    

    View Slide

46. Type #4
    Catalog organization
    Produces a set of result to present to users
    @galuhsahid
    

    View Slide

47. Recommender System
    Catalog Organization
    @galuhsahid
    

    View Slide

48. Type #5
    Generative model
    Focuses on generating data rather than classifying or organizing it
    @galuhsahid
    

    View Slide

49. Generating a new face
    Generative Model
    @galuhsahid
    

    View Slide

50. Generative Model
    Generating a new text
    @galuhsahid
    talktotransformer.com
    

    View Slide

51. Machine learning myths
    @galuhsahid
    

    View Slide

52. Myth #1
    Machine learning is neutral
    @galuhsahid
    

    View Slide

53. Myth #1
    Machine learning is neutral
    Machine learning systems are formulated and built by humans. Machine
    learning algorithms are trained by humans on data that are collected by
    humans and represent our (biased) current world.
    Since humans are biased, machine learning systems also inherit human’s
    biases.
    @galuhsahid
    

    View Slide

54. Gender-neutral translations
    Myth #1: Machine learning is neutral
    Source @galuhsahid
    

    View Slide

55. Gender-neutral translations
    Myth #1: Machine learning is neutral
    @galuhsahid
    

    View Slide

56. Gender-neutral translations
    Myth #1: Machine learning is neutral
    @galuhsahid
    

    View Slide

57. Text generation
    Myth #1: Machine learning is neutral
    Source @galuhsahid
    

    View Slide

58. Myth #2
    Machine learning is (or can be)
    perfect
    @galuhsahid
    

    View Slide

59. Myth #2
    Machine learning is (or can be)
    perfect
    Machine learning is never going to be perfect. Even humans can make errors.
    However, machine learning can improve.
    @galuhsahid
    

    View Slide

60. Feedback loop
    Myth #2: Machine learning is (or can be) perfect
    @galuhsahid
    

    View Slide

61. Feedback loop
    Myth #2: Machine learning is (or can be) perfect
    @galuhsahid
    

    View Slide

62. Feedback loop
    Myth #2: Machine learning is (or can be) perfect
    @galuhsahid
    

    View Slide

63. Feedback loop
    Myth #2: Machine learning is (or can be) perfect
    @galuhsahid
    

    View Slide

64. Myth #3
    We can just “build/apply ML” and
    call it a day
    @galuhsahid
    

    View Slide

65. Myth #3
    We can just “build/apply ML” and
    call it a day
    We still need to collaborate with others to ensure that we are solving the right
    problem. Machine learning is not going to figure out what the problem is.
    @galuhsahid
    

    View Slide

66. Example: recommendation system
    Myth #3: We can just “apply ML” and call it a day
    

    View Slide

67. Example: recommendation system
    Myth #3: We can just “apply ML” and call it a day
    @galuhsahid
    

    View Slide

68. Demo
    @galuhsahid
    

    View Slide

69. Rock Paper Scissors
    Demo
    @galuhsahid
    Code
    

    View Slide

70. Scroll with Your Voice
    Demo
    @galuhsahid
    Code
    

    View Slide

71. More machine learning
    • On building ML projects: First Steps Towards Your First Machine
    Learning Project
    • On ML with JavaScript: Machine Learning on the Web
    • On ML with TensorFlow: A Whirlwind Tour of Machine Learning
    with TensorFlow
    @galuhsahid
    

    View Slide

72. References
    • On UX and ML: Human-Centered Machine Learning (article)
    • On machine learning bias: Weapons of Math Destruction: How Big Data
    Increases Inequality and Threatens Democracy (book)
    • On end-to-end example of building ML-powered products: Building Machine
    Learning Powered Applications (book)
    @galuhsahid
    

    View Slide

73. Thank you!
    @galuhsahid
    

    View Slide