1
00:00:00,000 --> 00:00:04,544
[MUSIC]

2
00:00:04,544 --> 00:00:07,212
Now that we've discussed what
this course is going to be about,

3
00:00:07,212 --> 00:00:10,498
let's spend some time discussing what
we're going to assume your background

4
00:00:10,498 --> 00:00:13,470
knowledge is to successfully
complete this course.

5
00:00:13,470 --> 00:00:16,537
Well, there are a number of concepts
we're going to assume that you've seen in

6
00:00:16,537 --> 00:00:18,435
the previous courses in
this specialization.

7
00:00:18,435 --> 00:00:22,575
So, for example, in the foundations
course, we went through a lot of the high

8
00:00:22,575 --> 00:00:26,784
level concepts that we're going to present
in this specific about clustering and

9
00:00:26,784 --> 00:00:27,500
retrieval.

10
00:00:27,500 --> 00:00:31,095
But also, about how we think about machine
learning in terms of what are inputs and

11
00:00:31,095 --> 00:00:33,192
what are outputs in
machine learning methods?

12
00:00:33,192 --> 00:00:37,394
And how do we think about
analyzing these outputs?

13
00:00:37,394 --> 00:00:40,043
And we also built up
a lot of programming and

14
00:00:40,043 --> 00:00:44,340
data manipulation skills
in this foundations course.

15
00:00:44,340 --> 00:00:48,158
And then in the regression course we went
through a lot of detail on how we think

16
00:00:48,158 --> 00:00:51,181
about features representing
the inputs to our algorithms.

17
00:00:51,181 --> 00:00:54,785
And what the outputs are and
how to examine those outputs and

18
00:00:54,785 --> 00:00:57,607
assess different things about performance.

19
00:00:57,607 --> 00:01:01,111
And we talked about basical
statistical concepts like mean and

20
00:01:01,111 --> 00:01:04,363
variance that we're going to
see many times in this course.

21
00:01:04,363 --> 00:01:07,296
And also, we covered basic
machine learning concepts,

22
00:01:07,296 --> 00:01:09,689
like, what is a machine
learning algorithm?

23
00:01:09,689 --> 00:01:13,380
And then we went through specific
optimization algorithms that we're

24
00:01:13,380 --> 00:01:16,341
going to elude to in this course,
like coordinate ascent.

25
00:01:16,341 --> 00:01:21,272
And we talked about issues in terms of
model complexity, like overfitting and

26
00:01:21,272 --> 00:01:24,122
how to cope with that
using regularization.

27
00:01:24,122 --> 00:01:27,600
And these are all ideas that we're
going to refer to throughout this course.

28
00:01:27,600 --> 00:01:30,900
And we're going to assume that
you already know to some degree.

29
00:01:30,900 --> 00:01:34,772
Then, coming out of the classification
course, there are a number of things that

30
00:01:34,772 --> 00:01:37,802
are really crucial to the content
that we're going to cover here,

31
00:01:37,802 --> 00:01:40,554
like distributions and
conditional distributions, and

32
00:01:40,554 --> 00:01:42,203
maximum likelihood estimation.

33
00:01:42,203 --> 00:01:46,089
And then there are a couple things that
I'm going to refer to just to draw

34
00:01:46,089 --> 00:01:48,436
analogies, but aren't that critical,

35
00:01:48,436 --> 00:01:53,080
like ideas of linear classifiers,
multiclass classification, and boosting.

36
00:01:54,140 --> 00:01:57,593
So if you haven't taken the past
courses and some of these concepts,

37
00:01:57,593 --> 00:01:59,557
maybe you only have a fuzzy idea about,

38
00:01:59,557 --> 00:02:02,731
I'd strongly encourage you to
go back to these other courses.

39
00:02:02,731 --> 00:02:05,260
And at least watch the videos
related to these topics.

40
00:02:06,530 --> 00:02:11,019
Then, in terms of math background, there's
not going to be much of an emphasis on

41
00:02:11,019 --> 00:02:13,769
calculus like there was
in the past two courses.

42
00:02:13,769 --> 00:02:18,098
But we're still going to assume that you
know some basic linear algebra, like,

43
00:02:18,098 --> 00:02:23,100
what is a vector, what is a matrix, how
do you think about multiplying matrices?

44
00:02:23,100 --> 00:02:26,490
And I think that's basically it for
the linear algebra part.

45
00:02:26,490 --> 00:02:30,000
But in this course we're also going to
turn to some probabilistic concepts

46
00:02:30,000 --> 00:02:32,930
building off of things that you
saw in the classification course.

47
00:02:32,930 --> 00:02:36,410
So we're going to assume that you
know fundamental laws of probability,

48
00:02:36,410 --> 00:02:37,866
like probability sum to one.

49
00:02:37,866 --> 00:02:40,719
Or, they're bounded between zero and one.

50
00:02:40,719 --> 00:02:44,103
We're going to assume that you know,
what is a distribution, and

51
00:02:44,103 --> 00:02:46,872
how do you think about
a conditional distribution?

52
00:02:46,872 --> 00:02:49,313
And we're going to walk
through these concepts and

53
00:02:49,313 --> 00:02:52,897
teach them to you at the level that
you need to know them for this course.

54
00:02:52,897 --> 00:02:55,950
But it's definitely going to be helpful
if you have some of that background

55
00:02:55,950 --> 00:02:56,460
coming in.

56
00:02:58,010 --> 00:03:01,020
And in terms of programming experience,
we've tried to make this course as

57
00:03:01,020 --> 00:03:03,720
open as possible to people
having different preferences for

58
00:03:03,720 --> 00:03:05,580
different programming languages.

59
00:03:05,580 --> 00:03:11,240
But we're going to encourage people to
use Python although it's not required.

60
00:03:11,240 --> 00:03:12,050
For example,

61
00:03:12,050 --> 00:03:15,620
for all of the assignments we're
going to provide starter code in Python.

62
00:03:15,620 --> 00:03:18,830
So of course, if you're a Python user,
it's going to be easier for

63
00:03:18,830 --> 00:03:20,980
you to complete the assignments.

64
00:03:20,980 --> 00:03:23,950
But I want to emphasize that the point
of everything we're teaching in

65
00:03:23,950 --> 00:03:28,180
this course is to teach you fundamental
machine learning concepts and

66
00:03:28,180 --> 00:03:30,410
not specific implementation details.

67
00:03:30,410 --> 00:03:30,980
Though, of course,

68
00:03:30,980 --> 00:03:35,700
we want you to get hands on experience
with implementing the methods.

69
00:03:35,700 --> 00:03:38,410
But I want to say that if you
completed the foundations course, and

70
00:03:38,410 --> 00:03:41,380
hopefully the regression and
classification courses as well, you should

71
00:03:41,380 --> 00:03:46,540
be set for the skill level needed to
complete the assignmentss in this course.

72
00:03:46,540 --> 00:03:50,090
And if we think back to the foundations
course, we relied very heavily

73
00:03:50,090 --> 00:03:53,445
on pre-implemented algorithms like
those available on GraphLab Create.

74
00:03:53,445 --> 00:03:56,962
Because the point of that course was
just to understand the input and

75
00:03:56,962 --> 00:03:59,130
the output of machine learning algorithms.

76
00:03:59,130 --> 00:04:02,605
And we didn't get it to the details
of what's under the hood in each

77
00:04:02,605 --> 00:04:03,980
one of these algorithms.

78
00:04:03,980 --> 00:04:04,966
But this course,

79
00:04:04,966 --> 00:04:08,843
we're going to take a deep dive into
the algorithm in details, so much so

80
00:04:08,843 --> 00:04:13,720
that you should be able to implement
them in any language of your choice.

81
00:04:13,720 --> 00:04:17,000
And you're going to get practical
experience with doing this in this course.

82
00:04:17,000 --> 00:04:20,220
We're going to ask you to actually
implement these methods, for

83
00:04:20,220 --> 00:04:22,870
everything except in the last module,

84
00:04:22,870 --> 00:04:27,690
where we go through the more advanced
concept of latent Dirichlet allocation.

85
00:04:27,690 --> 00:04:31,500
Where actually we do teach you everything
you need to know to implement the methods,

86
00:04:31,500 --> 00:04:35,030
but we're not actually going to ask
you to do so in the assignment.

87
00:04:35,030 --> 00:04:38,040
So, to describe the assignments a little
bit more, we're going to follow the same

88
00:04:38,040 --> 00:04:41,430
structure that we did in the regression
and classification courses.

89
00:04:41,430 --> 00:04:44,840
Where in general, the structure is
going to to be that, first we're going to

90
00:04:44,840 --> 00:04:50,370
go through an exploration of the methods
using a preimplemented algorithm so

91
00:04:50,370 --> 00:04:53,220
that we can understand
the methods in more detail.

92
00:04:53,220 --> 00:04:56,650
And solidify the concepts
without getting bogged

93
00:04:56,650 --> 00:05:02,110
down in specific implementation details or
potential issues with bugs in our code.

94
00:05:02,110 --> 00:05:05,660
But once we've solidified these concepts,
then we're going to turn to

95
00:05:05,660 --> 00:05:07,830
actual implementations that
you're all going to write.

96
00:05:09,120 --> 00:05:12,160
And finally,
the computing needs are the same as in

97
00:05:12,160 --> 00:05:16,500
other courses of this specialization,
where there are a couple of choices.

98
00:05:16,500 --> 00:05:18,490
One is that you have your own machine.

99
00:05:18,490 --> 00:05:22,620
And if you're going to use SFrames,
which we encourage you to do to do

100
00:05:22,620 --> 00:05:27,700
the different data manipulations,
then you need to have a 64-bit machine.

101
00:05:27,700 --> 00:05:31,778
Other than that, it can be a fairly
basic desktop or laptop that you have.

102
00:05:31,778 --> 00:05:34,810
You're also ,of course,
going to need access to the Internet so

103
00:05:34,810 --> 00:05:38,900
that you can watch these lovely videos,
as well as download the assignments and

104
00:05:38,900 --> 00:05:42,440
upload your implementations
to the Coursera interface.

105
00:05:42,440 --> 00:05:45,250
And you're also going to need
the ability to install and

106
00:05:45,250 --> 00:05:49,150
run Python and
to store a few gigabytes of data.

107
00:05:49,150 --> 00:05:52,660
But, as an alternative, we're also
going to provide some preconfigured

108
00:05:52,660 --> 00:05:57,550
machines in the cloud, so that if you
don't have a 64-bit machine of your own,

109
00:05:57,550 --> 00:06:00,060
you can still complete this course.

110
00:06:00,060 --> 00:06:03,739
Okay, so now that we've gone through
what this course is about and

111
00:06:03,739 --> 00:06:08,091
everything you're going to need to
complete this course, let's get started.

112
00:06:08,091 --> 00:06:12,079
[MUSIC]