1
00:00:00,000 --> 00:00:04,213
[MUSIC]

2
00:00:04,213 --> 00:00:05,392
Hi.
I'm Carlos Guestrin.

3
00:00:05,392 --> 00:00:08,352
I'm the Amazon Professor
of Machine Learning.

4
00:00:08,352 --> 00:00:11,070
>> Wait, I'm the Amazon Professor
of Machine Learning.

5
00:00:11,070 --> 00:00:11,849
>> Who are you?

6
00:00:11,849 --> 00:00:13,253
>> I'm Emily Fox.

7
00:00:13,253 --> 00:00:16,500
>> Well, I'm the Amazon Professor of
Machine Learning in Computer Science.

8
00:00:16,500 --> 00:00:19,820
>> And I'm the Amazon Professor of
Machine Learning in Statistics.

9
00:00:19,820 --> 00:00:22,867
>> We're both at the University
of Washington in Seattle,

10
00:00:22,867 --> 00:00:24,329
this is the United States.

11
00:00:24,329 --> 00:00:25,979
[LAUGH] This is.

12
00:00:25,979 --> 00:00:28,980
>> This is the United States of America.

13
00:00:28,980 --> 00:00:29,500
>> Of America.

14
00:00:29,500 --> 00:00:30,060
>> To clarify.

15
00:00:30,060 --> 00:00:31,350
>> America.

16
00:00:31,350 --> 00:00:32,207
>> America.

17
00:00:32,207 --> 00:00:37,545
>> [LAUGH] So what are your
qualifications to be here today, Emily?

18
00:00:37,545 --> 00:00:40,430
[LAUGH]
>> Am I supposed to read my CV?

19
00:00:40,430 --> 00:00:44,220
No, okay, just the beginning of it.

20
00:00:44,220 --> 00:00:45,410
What are my qualifications?

21
00:00:45,410 --> 00:00:45,930
Okay.

22
00:00:45,930 --> 00:00:51,560
Well, I went to MIT for way too long,
nine years, though it wasn't all for

23
00:00:51,560 --> 00:00:55,105
one degree, got a couple of them
[LAUGH] during those nine years.

24
00:00:55,105 --> 00:00:57,310
But yeah, I was there for

25
00:00:57,310 --> 00:01:01,500
my PhD studying in electrical
engineering and computer science.

26
00:01:02,640 --> 00:01:04,110
Should I list more qualifications?

27
00:01:04,110 --> 00:01:07,580
I think that's pretty good.

28
00:01:07,580 --> 00:01:09,330
>> She's very well qualified.

29
00:01:09,330 --> 00:01:10,400
Let me just say that.

30
00:01:10,400 --> 00:01:17,420
[LAUGH] And I got my undergrad from
University of Sao Paulo in Brazil.

31
00:01:17,420 --> 00:01:20,240
And my PhD from Stanford
in computer science and

32
00:01:20,240 --> 00:01:24,410
I've been working in machine learning for
a long time, as Emily likes to say.

33
00:01:24,410 --> 00:01:25,950
>> A very long time.

34
00:01:25,950 --> 00:01:30,380
I guess maybe I should describe a little
bit about my transition to statistics.

35
00:01:30,380 --> 00:01:33,462
So, after I did my PhD in
electrical engineering and

36
00:01:33,462 --> 00:01:37,106
computer science,
I did a postdoc in statistics at Duke, and

37
00:01:37,106 --> 00:01:41,042
actually, honestly, a lot of my
PhD really was a statistics PhD.

38
00:01:41,042 --> 00:01:46,887
But MIT, as of the time of this recording,
does not have a statistics department,

39
00:01:46,887 --> 00:01:52,838
and did not at the time I studied there
so, yes, I'm a self-taught statistician.

40
00:01:54,670 --> 00:01:56,020
>> And we can all be self taught.

41
00:01:56,020 --> 00:01:56,992
>> Yes.
>> [LAUGH]

42
00:01:56,992 --> 00:01:59,210
>> That's what this MOOC is about.

43
00:01:59,210 --> 00:02:02,110
>> [LAUGH] Hey, we're learning together.

44
00:02:02,110 --> 00:02:06,763
So I've been working machine learning for
awhile and I'm really excited about

45
00:02:06,763 --> 00:02:10,246
methods in machine learning
from a wide range of settings.

46
00:02:10,246 --> 00:02:14,379
I've worked on theoretical algorithms for
what's called planning under uncertainty

47
00:02:14,379 --> 00:02:16,620
for making decisions
using machine learning.

48
00:02:16,620 --> 00:02:20,320
I've worked with sensor data
with distributed sensors,

49
00:02:20,320 --> 00:02:24,270
something called sensor networks which
today is called Internet of Things.

50
00:02:24,270 --> 00:02:30,837
But this is way back and then I took a big
focus on large-scale machine learning.

51
00:02:30,837 --> 00:02:34,777
So how do you scale up machine learning
methods to lots and lots and lots of data?

52
00:02:34,777 --> 00:02:38,890
And we'll use some of these ideas
in this specialization as well.

53
00:02:38,890 --> 00:02:40,020
How about you?

54
00:02:40,020 --> 00:02:42,950
>> I don't have quite as long
of a history as Carlos does.

55
00:02:42,950 --> 00:02:47,430
[LAUGH]
But the things that I really enjoy

56
00:02:47,430 --> 00:02:51,760
working on are time series applications,
the data that comes in over time and

57
00:02:51,760 --> 00:02:54,910
we want to find some underlying
structure or patterns to that data.

58
00:02:54,910 --> 00:02:58,498
And a lot of my focus in my research
is on how to scale up to really

59
00:02:58,498 --> 00:03:01,741
high dimensional or
large collections of time series,

60
00:03:01,741 --> 00:03:05,080
especially when they have
very complex dynamics.

61
00:03:05,080 --> 00:03:08,140
And another one of my
big interests is how to

62
00:03:08,140 --> 00:03:11,890
make inferences online
as data is streaming in.

63
00:03:11,890 --> 00:03:15,100
And I've worked on a wide range
of different applications,

64
00:03:15,100 --> 00:03:17,550
some that I'm really passionate about.

65
00:03:17,550 --> 00:03:20,978
One in particular, right now,
is neuroscience, but

66
00:03:20,978 --> 00:03:25,152
I've also worked on speech data,
motion capture applications,

67
00:03:25,152 --> 00:03:29,643
analyzing human motion, to financial
data and the list goes on and on.

68
00:03:29,643 --> 00:03:34,193
But lots of data have really
important temporal structure and

69
00:03:34,193 --> 00:03:37,400
that's really a passion of mine.

70
00:03:37,400 --> 00:03:38,320
>> Yeah.
>> No, yeah, that's it.

71
00:03:38,320 --> 00:03:40,298
>> And we've been teaching machine
learning for a long time and

72
00:03:40,298 --> 00:03:43,870
maybe in the university setting-
>> Some of us longer than others.

73
00:03:43,870 --> 00:03:45,247
>> This joke never ends.

74
00:03:45,247 --> 00:03:48,660
>> No.
>> And many at the university setting.

75
00:03:48,660 --> 00:03:51,960
So here we are,
it's an exciting experience for

76
00:03:51,960 --> 00:03:55,015
us to teach in this online course.

77
00:03:55,015 --> 00:03:55,567
>> Yes.

78
00:03:55,567 --> 00:03:59,403
>> So how did you decide
to do this online course?

79
00:03:59,403 --> 00:04:00,900
>> How did I decide?

80
00:04:00,900 --> 00:04:02,230
That's a very good question.

81
00:04:04,160 --> 00:04:08,010
Well, no, there are a lot of reasons
why I wanted to do this online course.

82
00:04:08,010 --> 00:04:12,570
But really, the most fundamental reason
was thinking about how to disseminate

83
00:04:12,570 --> 00:04:14,480
machine learning to a very broad audience,

84
00:04:14,480 --> 00:04:18,440
because personally, I came from a very
different background and training.

85
00:04:18,440 --> 00:04:24,170
And actually Carlos did as well and
so I, myself, have

86
00:04:24,170 --> 00:04:28,340
gone through this experience of trying to
learn machine learning, learn statistics.

87
00:04:28,340 --> 00:04:33,109
And I think there's a really nice way that
it can be taught at a very fundamental

88
00:04:33,109 --> 00:04:36,879
level to a very broad audience and
that's one of my passions.

89
00:04:36,879 --> 00:04:42,792
And another reason is I'd like to think
about this for other goals in the future

90
00:04:42,792 --> 00:04:48,614
of how to convey this to other groups that
I interact with on campus as a teacher

91
00:04:48,614 --> 00:04:53,658
at the University of Washington,
professor [LAUGH], assistant.

92
00:04:53,658 --> 00:04:55,619
[LAUGH]
>> Not for long.

93
00:04:55,619 --> 00:05:01,940
So, yeah, and I'm very excited about
how machine learning gets taught.

94
00:05:01,940 --> 00:05:05,890
And we really have been thinking
a lot about how to approach

95
00:05:05,890 --> 00:05:08,800
this really broad idea of
machine learning in a new way.

96
00:05:08,800 --> 00:05:11,390
So in this specialization,

97
00:05:11,390 --> 00:05:15,160
as we'll talk more about in this module,
we'll experience machine learning

98
00:05:15,160 --> 00:05:18,260
in a whole different way than
you experienced in other places.

99
00:05:18,260 --> 00:05:21,410
We're really gonna be focused on for
applications first,

100
00:05:21,410 --> 00:05:24,890
understanding how machine learning
has impact and then digging in and

101
00:05:24,890 --> 00:05:28,080
understanding how those methods
are built and how they can be useful.

102
00:05:28,080 --> 00:05:29,620
So part of the journey for

103
00:05:29,620 --> 00:05:34,260
us is to build a new way to teach
machine learning, which is cool.

104
00:05:34,260 --> 00:05:35,090
>> It is cool.

105
00:05:35,090 --> 00:05:35,600
>> So we're excited.

106
00:05:35,600 --> 00:05:37,177
>> Yes.

107
00:05:37,177 --> 00:05:38,234
We're excited.

108
00:05:38,234 --> 00:05:39,950
[LAUGH]

109
00:05:39,950 --> 00:05:43,629
[MUSIC]