1
00:00:00,000 --> 00:00:04,262
Welcome to this free online class on
machine learning. Machine learning is one

2
00:00:04,262 --> 00:00:08,579
of the most exciting recent technologies.
And in this class, you learn about the

3
00:00:08,579 --> 00:00:13,115
state of the art and also gain practice
implementing and deploying these algorithms

4
00:00:13,115 --> 00:00:17,487
yourself. You've probably use a learning
algorithm dozens of times a day without

5
00:00:17,487 --> 00:00:21,422
knowing it. Every time you use a web
search engine like Google or Bing to

6
00:00:21,422 --> 00:00:25,794
search the internet, one of the reasons
that works so well is because a learning

7
00:00:25,794 --> 00:00:30,002
algorithm, one implemented by Google or
Microsoft, has learned how to rank web

8
00:00:30,002 --> 00:00:35,144
pages. Every time you use Facebook or
Apple's photo typing application and it

9
00:00:35,144 --> 00:00:40,595
recognizes your friends' photos, that's
also machine learning. Every time you read

10
00:00:40,595 --> 00:00:46,054
your email and your spam filter saves you
from having to wade through tons of spam

11
00:00:46,054 --> 00:00:50,980
email, that's also a learning algorithm.
For me one of the reasons I'm excited is

12
00:00:50,980 --> 00:00:55,643
the AI dream of someday building machines
as intelligent as you or me. We're a long

13
00:00:55,643 --> 00:01:00,076
way away from that goal, but many AI
researchers believe that the best way to

14
00:01:00,076 --> 00:01:04,567
towards that goal is through learning
algorithms that try to mimic how the human

15
00:01:04,567 --> 00:01:08,994
brain learns. I'll tell you a little bit
about that too in this class. In this

16
00:01:08,994 --> 00:01:13,542
class you learn about state-of-the-art
machine learning algorithms. But it turns

17
00:01:13,542 --> 00:01:17,919
out just knowing the algorithms and
knowing the math isn't that much good if

18
00:01:17,919 --> 00:01:22,466
you don't also know how to actually get
this stuff to work on problems that you

19
00:01:22,466 --> 00:01:26,844
care about. So, we've also spent a lot
of time developing exercises for you to

20
00:01:26,844 --> 00:01:32,088
implement each of these algorithms and
see how they work fot yourself. So why is

21
00:01:32,088 --> 00:01:37,075
machine learning so prevalent today?
It turns out that machine learning is a

22
00:01:37,075 --> 00:01:41,713
field that had grown out of the field of
AI, or artificial intelligence. We wanted

23
00:01:41,713 --> 00:01:46,642
to build intelligent machines and it turns
out that there are a few basic things that

24
00:01:46,642 --> 00:01:51,454
we could program a machine to do such as
how to find the shortest path from A to B.

25
00:01:51,454 --> 00:01:56,267
But for the most part we just did not know
how to write AI programs to do the more

26
00:01:56,267 --> 00:02:00,905
interesting things such as web search or
photo tagging or email anti-spam. There

27
00:02:00,905 --> 00:02:05,718
was a realization that the only way to do
these things was to have a machine learn

28
00:02:05,718 --> 00:02:11,237
to do it by itself. So, machine learning
was developed as a new capability for

29
00:02:11,237 --> 00:02:16,950
computers and today it touches many
segments of industry and basic science.

30
00:02:16,950 --> 00:02:21,496
For me, I work on machine learning and
in a typical week I might end up talking to

31
00:02:21,496 --> 00:02:25,698
helicopter pilots, biologists, a bunch
of computer systems people (so my

32
00:02:25,698 --> 00:02:30,590
colleagues here at Stanford) and averaging
two or three times a week I get email from

33
00:02:30,590 --> 00:02:35,021
people in industry from Silicon Valley
contacting me who have an interest in

34
00:02:35,021 --> 00:02:39,741
applying learning algorithms to their own
problems. This is a sign of the range of

35
00:02:39,741 --> 00:02:44,000
problems that machine learning touches.
There is autonomous robotics, computational

36
00:02:44,000 --> 00:02:48,777
biology, tons of things in Silicon Valley
that machine learning is having an impact

37
00:02:48,777 --> 00:02:55,320
on. Here are some other examples of
machine learning. There's database mining.

38
00:02:55,320 --> 00:03:00,063
One of the reasons machine learning has so
pervaded is the growth of the web and the

39
00:03:00,063 --> 00:03:04,751
growth of automation All this means that
we have much larger data sets than ever

40
00:03:04,751 --> 00:03:09,272
before. So, for example tons of Silicon
Valley companies are today collecting web

41
00:03:09,272 --> 00:03:13,737
click data, also called clickstream data,
and are trying to use machine learning

42
00:03:13,737 --> 00:03:18,481
algorithms to mine this data to understand
the users better and to serve the users

43
00:03:18,481 --> 00:03:22,327
better, that's a huge segment of
Silicon Valley right now. Medical

44
00:03:22,327 --> 00:03:27,483
records. With the advent of automation, we
now have electronic medical records, so if

45
00:03:27,483 --> 00:03:32,640
we can turn medical records into medical
knowledge, then we can start to understand

46
00:03:32,640 --> 00:03:37,238
disease better. Computational biology.
With automation again, biologists are

47
00:03:37,238 --> 00:03:41,774
collecting lots of data about gene
sequences, DNA sequences, and so on, and

48
00:03:41,774 --> 00:03:46,931
machines running algorithms are giving us
a much better understanding of the human

49
00:03:46,931 --> 00:03:51,376
genome, and what it means to be human.
And in engineering as well, in all fields of

50
00:03:51,376 --> 00:03:55,034
engineering, we have larger and larger,
and larger and larger data sets, that

51
00:03:55,034 --> 00:03:59,249
we're trying to understand using learning
algorithms. A second range of machinery

52
00:03:59,249 --> 00:04:03,440
applications is ones that we cannot
program by hand. So for example, I've

53
00:04:03,440 --> 00:04:08,328
worked on autonomous helicopters for many
years. We just did not know how to write a

54
00:04:08,328 --> 00:04:18,023
computer program to make this helicopter
fly by itself. The only thing that worked

55
00:04:18,023 --> 00:04:35,580
was having a computer learn by itself how
to fly this helicopter. [Helicopter whirling]

56
00:04:37,120 --> 00:04:42,880
Handwriting recognition. It turns out one
of the reasons it's so inexpensive today to

57
00:04:42,880 --> 00:04:47,330
route a piece of mail across the
countries, in the US and internationally,

58
00:04:47,330 --> 00:04:51,899
is that when you write an envelope like
this, it turns out there's a learning

59
00:04:51,899 --> 00:04:56,943
algorithm that has learned how to read your
handwriting so that it can automatically

60
00:04:56,943 --> 00:05:01,749
route this envelope on its way, and so it
costs us a few cents to send this thing

61
00:05:01,749 --> 00:05:06,318
thousands of miles. And in fact if you've
seen the fields of natural language

62
00:05:06,318 --> 00:05:10,531
processing or computer vision,
these are the fields of AI pertaining to

63
00:05:10,531 --> 00:05:15,321
understanding language or understanding
images. Most of natural language processing

64
00:05:15,321 --> 00:05:20,689
and most of computer vision today is
applied machine learning. Learning

65
00:05:20,689 --> 00:05:25,576
algorithms are also widely used for self-
customizing programs. Every time you go to

66
00:05:25,576 --> 00:05:30,286
Amazon or Netflix or iTunes Genius, and it
recommends the movies or products and

67
00:05:30,286 --> 00:05:35,073
music to you, that's a learning algorithm.
If you think about it they have million

68
00:05:35,073 --> 00:05:39,999
users; there is no way to write a million
different programs for your million users.

69
00:05:39,999 --> 00:05:44,807
The only way to have software give these
customized recommendations is to become

70
00:05:44,807 --> 00:05:49,258
learn by itself to customize itself to
your preferences. Finally learning

71
00:05:49,258 --> 00:05:53,294
algorithms are being used today to
understand human learning and to

72
00:05:53,294 --> 00:05:58,042
understand the brain. We'll talk about
how researches are using this to make

73
00:05:58,042 --> 00:06:03,182
progress towards the big AI dream. A few
months ago, a student showed me an article

74
00:06:03,182 --> 00:06:07,996
on the top twelve IT skills. The skills
that information technology hiring

75
00:06:07,996 --> 00:06:13,006
managers cannot say no to. It was a
slightly older article, but at the top of

76
00:06:13,006 --> 00:06:17,988
this list of the twelve most desirable IT
skills was machine learning. Here at

77
00:06:17,988 --> 00:06:21,793
Stanford, the number of recruiters
that contact me asking if I know any

78
00:06:21,793 --> 00:06:25,920
graduating machine learning students
is far larger than the machine learning

79
00:06:25,920 --> 00:06:30,047
students we graduate each year. So I
think there is a vast, unfulfilled demand

80
00:06:30,047 --> 00:06:34,280
for this skill set, and this is a great time to
be learning about machine learning, and I

81
00:06:34,280 --> 00:06:38,454
hope to teach you a lot about machine
learning in this class. In the next video,

82
00:06:38,454 --> 00:06:42,123
we'll start to give a more formal
definition of what is machine learning.

83
00:06:42,123 --> 00:06:46,044
And we'll begin to talk about the main
types of machine learning problems and

84
00:06:46,044 --> 00:06:49,864
algorithms. You'll pick up some of the
main machine learning terminology, and

85
00:06:49,864 --> 00:06:53,684
start to get a sense of what are the
different algorithms, and when each one

86
00:06:53,684 --> 00:06:54,740
might be appropriate.