Introduction ML

1. Introduction to
Machine Learning
Danna Gurari
University of Texas at Austin
Spring 2021
https://www.ischool.utexas.edu/~dannag/Courses/IntroToMachineLearning/CourseContent.html

2. Today’s Topics
• Machine learning applications
• History of machine learning
• How does a machine learn?
• Class logistics
• Lab

3. Today’s Topics
• Machine learning applications
• History of machine learning
• How does a machine learn?
• Class logistics
• Lab

4. Key Motivations for Machine Learning
Systems that support humans by either
improving upon existing human capabilities
or providing new capabilities

5. Problems Solved by Machine Learning Today
Spam Detection

6. Problems Solved by Machine Learning Today
Information Retrieval

7. Problems Solved by Machine Learning Today
Recognition
(Face) (Speech) (Fraud)

8. Problems Solved by Machine Learning Today
Robotics
(Self-driving Vehicles) (Medical Surgery) (Manufacturing)

9. Problems Solved by Machine Learning Today
Recommendation Systems

10. Problems Solved by Machine Learning Today
e.g., recognizing people
e.g., shopping without a cashier
e.g., self-driving vehicle on Mars
Computer Vision Systems

11. Problems Solved by Machine Learning Today
e.g., Amazon’s Echo with Alexa e.g., Google Home
Home Virtual Assistants

12. Today’s Topics
• Machine learning applications
• History of machine learning
• How does a machine learn?
• Class logistics
• Lab

13. Origins of ML: Scaling Human Abilities
1613
Human “Computers”: first reference to people who perform calculations towards solving complex problems
http://whencomputerswerehuman.djaghe.com/

14. Origins of ML: Scaling Human Abilities
1613
• e.g., supported NASA space travel in early 1960s
Dorothy Vaughn Mary Jackson Miriam Mann
Excellent summary: https://en.wikipedia.org/wiki/Human_computer
Human “Computers”: first reference to people who perform calculations towards solving complex problems

15. Origins of ML: Scaling Human Abilities
1613
Human “Computers”
1945
ENIAC (Electronic Numerical Integrator and
Computer) created during World War II
(could compute 5,000 additions in one second)
First programmable machine
Human computers became first programmers

16. Origins of ML: Conceptual Framework
1613
Human “Computers”
1945 1950
Turing Test: can ”C” decide whether text
responses come from a machine or human
Turing Test
First programmable machine
Alan Turing
(1912-1954)

17. Origins of ML: Conceptual Framework
1613
Human “Computers”
1945 1956
Artificial
Intelligence
First programmable machine
“Artificial intelligence” established as a field at a workshop

18. Origins of ML: Conceptual Framework
1613
Human “Computers”
1945 1956
Artificial
Intelligence
First programmable machine
“Artificial intelligence” established as a field at a workshop
Workshop Proposal: “… We propose that a 2 month, 10 man study of artificial intelligence be
carried out during the summer of 1956 at Dartmouth College in Hanover, New Hampshire. The
study is to proceed on the basis of the conjecture that every aspect of learning or any other
feature of intelligence can in principle be so precisely described that a machine can be made
to simulate it. An attempt will be made to find how to make machines use language, form
abstractions and concepts, solve kinds of problems now reserved for humans, and improve
themselves. We think that a significant advance can be made in one or more of these
problems if a carefully selected group of scientists work on it together for a summer…”

19. Origins of ML: Conceptual Framework
1613
Human “Computers”
1945
First programmable machine Turing Test
& Artificial
Intelligence
1959
AI researcher Arthur Samuel coins the term
“machine learning” as:
“Field of study that gives computers the ability
to learn without being explicitly programmed.”
Machine
Learning
Artificial Intelligence
(machines that do
“intelligent” things)
Machine Learning
(algorithms that “learn”
for themselves)
1956

20. Motivation for Machines that “Learn”
• Process for hand-crafted rules:
Source: https://www.oreilly.com/library/view/hands-on-machine-learning/9781491962282/ch01.html

21. Motivation for Machines that “Learn”: Class Task
e.g., What rules would you use to answer: “Is a person in the image?”

22. Motivation for Machines that “Learn”
e.g., are these lines parallel?

23. Motivation for Machines that “Learn”
e.g., are these lines parallel?

24. Motivation for Machines that “Learn”
1. It is hard to hand-craft a complete set of rules
2. We, as humans, may not devise the best rules for a machine since our brains
(unconsciously) pre-process the data we sense

25. Motivation for Machines that “Learn”
Should you design rules or use machine learning for these tasks:
• Count how many times the letter “F” shows up in this sentence: FINISHED FILES ARE THE
RESULT OF YEARS OF SCIENTIFIC STUDY COMBINED WITH THE EXPERIENCE OF YEARS?
• Calculate the cost for gasoline on a road trip?

26. Origins of ML: Rises and Falls of AI/ML Popularity
1613
Human “Computers”
1945
First programmable machine Turing Test
& Artificial
Intelligence
1959
Machine
Learning
1956
Wave 1 Wave 2
(according
to
Google
Books)
Ian Goodfellow, Yoshua Bengio, and Aaron Courville; Deep Learning, 2016
1974 1980 1987 1993
1rst AI
Winter
2nd AI
Winter

27. Origins of ML: Rises and Falls of AI/ML Popularity
1613
Human “Computers”
1945
First programmable machine Turing Test
& Artificial
Intelligence
1959
Machine
Learning
1956 2006
1974 1980 1987 1993
When will be the next fall of AI and ML?
1rst AI
Winter
2nd AI
Winter

28. Today’s Topics
• Machine learning applications
• History of machine learning
• How does a machine learn?
• Class logistics
• Lab

29. General Idea
Excellent reference: https://machinelearningmastery.com/difference-between-algorithm-and-model-in-machine-learning/
An algorithm learns from data
patterns that a final model will
use to make a prediction

30. General Idea
An algorithm learns from data
patterns that a final model will
use to make a prediction

31. Data Types: What a Machine Learns From?
• Audio
• Input?
e.g.,

32. Data Types: What a Machine Learns From?
• Audio
• Input?
• Images
• Input?

33. Data Types: What a Machine Learns From?
• Audio
• Input?
• Images
• Input?
• Video
• Input?
Time 1
1 hour
Analogous to:

34. Data Types: What a Machine Learns From?
• Audio
• Input?
• Images
• Input?
• Video
• Input?
• Text
• Input?
e.g.,

35. Data Types: What a Machine Learns From?
• Audio
• Input?
• Images
• Input?
• Video
• Input?
• Text
• Input?
• Multi-modal
• Input? - combination of the above

36. Data Types: Many Public Datasets Available
• Dataset creation is beyond the scope of this class
• We will benefit from other people’s efforts:
• Google Dataset Search
• Amazon’s AWS datasets
• Kaggle datasets
• Wikipedia’s list
• UC Irvine Machine Learning Repository
• Quora.com
• Reddit
• Dataportals.org
• Opendatamonitor.eu
• Quandl.com

37. General Idea
An algorithm learns from data
patterns that a final model will
use to make a prediction

38. • Unsupervised
• Discover patterns/structures
in the data
How to Learn?
• Supervised
• Learn to predict for novel cases
by studying correct outputs for
many data points

39. How to Learn?
• Unsupervised
• No label given for training data
• Supervised
• Label given for training data: e.g., “cat”
What is this?

40. How to Learn?
• Unsupervised
• No label given for training data
• Supervised
• Label given for training data: e.g., “berimbau”
What is this?

41. How to Learn?
• Unsupervised
• No label given for training data
• Supervised
• Label given for training data: e.g., “yes”
Is this email spam?

42. Types of “Unsupervised” Learning Tasks
Hands-on Machine Learning with Scikit-Learn & TensorFlow, Aurelien Geron
Clustering Anomaly Detection
What are real world applications for these types?

43. Types of “Supervised” Learning Tasks
Hands-on Machine Learning with Scikit-Learn & TensorFlow, Aurelien Geron
Regression
(predict continuous value)
Classification
(predict discrete value)
What are real world applications for these types?

44. Supervised Learning: How to Teach a Machine?
Instance-Based Learning
Hands-on Machine Learning with Scikit-Learn & TensorFlow, Aurelien Geron

45. Supervised Learning: How to Teach a Machine?
Model-Based Learning
Hands-on Machine Learning with Scikit-Learn & TensorFlow, Aurelien Geron

46. Supervised Learning: How to Teach a Machine?
Model-Based Learning
• Goal: learn data distribution in the “real world”
• Task: create increasingly complex models to separate x from o
• e.g., simple = linear
• e.g., more complex = quadratic
Figure source: https://medium.com/greyatom/what-is-underfitting-and-
overfitting-in-machine-learning-and-how-to-deal-with-it-6803a989c76

47. Supervised Learning: How to Teach a Machine?
Model-Based Learning
• Goal: learn data distribution in the “real world”
• Modeling: increase vs decrease model’s representational capacity
Figure source: https://medium.com/greyatom/what-is-underfitting-and-
overfitting-in-machine-learning-and-how-to-deal-with-it-6803a989c76

48. Supervised Learning: How to Teach a Machine?
Online Learning
Hands-on Machine Learning with Scikit-Learn & TensorFlow, Aurelien Geron

49. Supervised Learning: How to Teach a Machine?
Online Learning
Hands-on Machine Learning with Scikit-Learn & TensorFlow, Aurelien Geron
Why learn incrementally?

50. Supervised Learning: How to Teach a Machine?
Offline Learning
Hands-on Machine Learning with Scikit-Learn & TensorFlow, Aurelien Geron
Algorithm cannot
learn incrementally

51. General Idea
An algorithm learns from data
patterns that a final model will
use to make a prediction

52. Algorithm Scope for Class:
Last 61 Years And More
1613
Human “Computers”
1945
First programmable machine Turing Test
& Artificial
Intelligence
1959
Machine
Learning
1956 2006
1974 1980 1987 1993
1rst AI
Winter
2nd AI
Winter

53. Algorithm Scope: Next 5 Lectures
e.g., Linear
Regression,
Decision Tree,
Naïve Bayes,
KNN, SVM,
Boosting,
Bagging,
Stacking
Ian Goodfellow, Yoshua Bengio, and Aaron Courville; Deep Learning, 2016.

54. Algorithm Scope: Middle 4 lectures
Ian Goodfellow, Yoshua Bengio, and Aaron Courville; Deep Learning, 2016.

55. Algorithm Scope: Other Topics
Ian Goodfellow, Yoshua Bengio, and Aaron Courville; Deep Learning, 2016.

56. Putting It All Together
An algorithm learns from data
patterns that a final model will
use to make a prediction

57. Putting It All Together: Analogous to a Love Story of
Partnering Up and Road Tripping Somewhere
An algorithm learns from data
patterns that a final model will
use to make a prediction

58. Putting It All Together: Analogous to a Love Story of
Partnering Up and Road Tripping Somewhere
Key Issue: How Fast Will You Get There?
(more on this when we discuss CPU and GPU hardware)

59. Putting It All Together: Analogous to a Love Story of
Partnering Up and Road Tripping Somewhere
Key Issue: Where Will You Go?

60. Putting It All Together: Where Will You Go?
https://www.theverge.com/2015/7/1/8880363/google-
apologizes-photos-app-tags-two-black-people-gorillas

61. Putting It All Together: Where Will You Go?
https://www.theverge.com/2015/7/1/8880363/google-
apologizes-photos-app-tags-two-black-people-gorillas
Why do you think the
algorithm made this mistake?

62. Putting It All Together: Where Will You Go?
Two kids bought their
mom a Nikon Coolpix
S630 digital camera for
Mother's Day… when
they took portrait
pictures of each other, a
message flashed across
the screen asking, "Did
someone blink?"
http://content.time.com/time/business/article/0,8599,1954643,00.html

63. Putting It All Together: Where Will You Go?
http://content.time.com/time/business/article/0,8599,1954643,00.html
Why do you think the
algorithm made this mistake?

64. Putting It All Together: Where Will You Go?
Algorithm identifies men in kitchens as women. Learned this example
from given dataset. (Zhao, Wang, Yatskar, Ordonez, Chang, 2017)
https://www.wired.com/story/machines-taught-by-photos-learn-a-sexist-view-of-women/

65. Putting It All Together: Where Will You Go?
https://www.wired.com/story/machines-taught-by-photos-learn-a-sexist-view-of-women/
Algorithm identifies men in kitchens as women. Learned this example
from given dataset. (Zhao, Wang, Yatskar, Ordonez, Chang, 2017)
Why do you think the
algorithm made this mistake?

66. Today’s Topics
• Machine learning applications
• History of machine learning
• How does a machine learn?
• Class logistics
• Lab

67. Introductions
Instructor: Danna Gurari
Danna: pronounced like “Donna”
Gurari: rhymes with Ferrari
Interdisciplinary class: share your (1) name, (2) preferred pronouns, and (3) career goal
Teaching Assistant: Pei-Chih “Patrick” Chao
Office hours: Mon 3-5pm, Thurs 4-5pm
Email address: pchao@utexas.edu

68. Introductions
NameCoach: a way to share your
name pronunciation in Canvas
To record your name:
1. Find NameCoach in Canvas
courses page
2. Click on record button to
start
3. Check your recording by
clicking on play button
1. NameCoach 3. Play 2. Record/Edit

69. Course Objectives
• Understand the key concepts in machine learning:
1. Characterize the process to train and test machine learning algorithms
2. Identify the challenges for designing modern machine learning algorithms
that can harness today’s “big” datasets
3. Recognize the strengths and weaknesses of different ways to evaluate
machine learning algorithms
4. Critique core and cutting edge machine learning algorithms

70. Course Objectives
• Apply machine learning systems to perform various AI tasks:
1. Develop programming skills by writing code in Python
2. Experiment with machine learning libraries, including scikit-learn and Keras
3. Evaluate machine learning algorithms for tasks in various application
domains, including for analyzing text and analyzing images
4. Employ cloud computing resources in order to take advantage of modern
hardware and software platforms

71. Course Objectives
• Conduct and communicate original research:
1. Propose a novel research idea (this will be an iterative process)
2. Design and execute experiments to support the proposed idea
3. Write a research paper about the project (and possibly submit it for
publication)
4. Present the project to the class

72. Class Overview
• Class website
• https://www.ischool.utexas.edu/~dannag/Courses/IntroToMachineLearning/
• Class objectives, schedule, assignments, and policies
• https://www.ischool.utexas.edu/~dannag/Courses/IntroToMachineLearning/Syllabus
/Syllabus.pdf
• Grading (from class syllabus):

73. Q&A: “What are the assignments?”
• 5 problem sets (first assignment due next week)
• 3 programming assignments
• Final project
• Pre-proposal
• Proposal
• Outline
• Video Presentation
• Peer evaluation
• Final project submission (final report, code, and video)
• Late policy
• Penalized 1% of grade per hour for up to 8 hours
• No credit if more than 8 hours late

74. Q&A: “Do I have the appropriate
pre-requisites/background?”
• Yes. While there are no pre-requisites, programming
experience is strongly recommended.
• You will be expected to further develop skills we cover in class
on your own
• Programming; e.g., Python
• Linear algebra; e.g., vector/matrix manipulations
• Calculus; e.g., partial derivatives
• Probability; e.g., Bayes rule

75. Q&A: “What are required textbooks?”
Required Strongly recommended

76. Class Format
• Mondays = lecture & group discussions
• Tuesdays = recorded in-class lab tutorial shared by 10am followed by
open, optional Q&A session from 4-5pm

77. Congratulations!
• By taking this class, you receive a gift of:
• Thanks to:

78. What is My “Why” for Teaching You…
WHY?
To guide and witness you
discover more about
your potential and your
passions
HOW?
By empowering you to become
proficient in one of my passions
WHAT?
Machine Learning

79. Today’s Topics
• Class logistics
• Machine learning applications
• History of machine learning
• How does a machine learn?
• Lab

80. Figure Credits
• https://www.amazon.com/Hands-Machine-Learning-Scikit-Learn-TensorFlow/dp/9352135210
• https://www.amazon.com/Make-Your-Own-Neural-Network-ebook/dp/B01EER4Z4G
• https://www.amazon.com/Deep-Learning-Adaptive-Computation-
Machine/dp/0262035618/ref=sr_1_1?ie=UTF8&qid=1472485235&sr=8-1&keywords=deep+learning+book
• https://www.amazon.com/Hidden-Figures-Taraji-P-Henson/dp/B01MU84AWP
• https://www.nasa.gov/content/dorothy-vaughan-biography
• https://interestingengineering.com/mary-jackson-remembering-nasas-first-black-female-engineer
• https://www.bbc.com/news/magazine-39003904
• http://www.pimall.com/nais/pivintage/enic.html
• https://en.wikipedia.org/wiki/Turing_test
• https://en.wikipedia.org/wiki/Alan_Turing
• https://pdfs.semanticscholar.org/d486/9863b5da0fa4ff5707fa972c6e1dc92474f6.pdf
• http://brainden.com/line-illusions.htm
• https://www.pregnancybirthbaby.org.au/learning-to-crawl
• https://www.parents.com/baby/development/intellectual/age-by-age-guide-to-reading-to-your-baby/
• https://www.worthpoint.com/worthopedia/ms-2003-ceramic-purple-car-1861052763