1
00:00:00,726 --> 00:00:04,268
[MUSIC]

2
00:00:04,268 --> 00:00:08,076
In this specialization we're going
to use a number of Python tools that

3
00:00:08,076 --> 00:00:10,995
really help us get started
with machine learning, and

4
00:00:10,995 --> 00:00:14,850
build applications that
really scale to lots of data.

5
00:00:14,850 --> 00:00:17,510
One of them is called SFrames,

6
00:00:17,510 --> 00:00:23,310
which is a really scalable data structure
for dealing with big tables of data.

7
00:00:23,310 --> 00:00:27,090
It doesn't have to fit in memory, it goes
into the disk of your computer and so

8
00:00:27,090 --> 00:00:29,780
can scale up to millions of rows of data,

9
00:00:29,780 --> 00:00:31,510
even if you don't have that
much memory in your machine.

10
00:00:31,510 --> 00:00:34,030
And we'll see lots of examples of that.

11
00:00:34,030 --> 00:00:38,804
And that data search with SFrame is
part of a package called GraphLab Create

12
00:00:38,804 --> 00:00:41,770
that we'll also use in
this specialization.

13
00:00:42,790 --> 00:00:48,090
So let's hear an example from IPython
notebook of how to use the SFrame.

14
00:00:49,800 --> 00:00:55,540
So here's a IPython notebook, just
like we did with the previous example.

15
00:00:55,540 --> 00:00:59,611
And I'm gonna call it,

16
00:00:59,611 --> 00:01:05,115
Getting Started with SFrames.

17
00:01:05,115 --> 00:01:09,492
And just to get a little
bit more space on my slide,

18
00:01:09,492 --> 00:01:14,090
I'm going to hide that header and
hide the toolbar.

19
00:01:14,090 --> 00:01:17,460
So now there's a little bit
more space in the slide.

20
00:01:17,460 --> 00:01:21,240
Now to get those frames you have
to fire up GraphLab Create, and so

21
00:01:21,240 --> 00:01:24,060
firing that up is pretty easy.

22
00:01:24,060 --> 00:01:30,111
So I'm hitting ask+M here to get
a textbox just like a wiki page and

23
00:01:30,111 --> 00:01:36,480
I'm going to fire up, oops,
that, I hit enter too quickly.

24
00:01:36,480 --> 00:01:38,220
I'm going, let me just edit that again.

25
00:01:38,220 --> 00:01:42,250
It does not wanna let me edit it.

26
00:01:42,250 --> 00:01:48,950
There, fire up GraphLab Create.

27
00:01:48,950 --> 00:01:52,310
So every time we wanna start
GraphLab Create, all we have to do.

28
00:01:52,310 --> 00:01:54,940
After we start, is type import graphlab.

29
00:01:54,940 --> 00:01:58,663
And now we have all the tools of
GraphLab Create available for us,

30
00:01:58,663 --> 00:02:03,960
including the SFrame, and any algorithms
that we'll use throughout this course.

31
00:02:05,650 --> 00:02:11,920
So, the first thing that I'm going
to do is load some data from disk.

32
00:02:11,920 --> 00:02:20,024
So, I'm going to #Load a tabular data set.

33
00:02:20,024 --> 00:02:24,750
So, loading a tabular data set
as an SFrame is pretty simple.

34
00:02:24,750 --> 00:02:29,080
The data set can be in
many different formats.

35
00:02:29,080 --> 00:02:32,350
What we're gonna use is what's
called a CSV format in this example,

36
00:02:32,350 --> 00:02:34,800
which is just a comma separated file.

37
00:02:34,800 --> 00:02:36,950
And that file is in my current directory.

38
00:02:36,950 --> 00:02:42,382
And actually,
I'm gonna create a variable in SFrame and

39
00:02:42,382 --> 00:02:46,170
just call it sf for this example.

40
00:02:46,170 --> 00:02:50,230
All I have to do is graphlab.SFrame and

41
00:02:50,230 --> 00:02:54,590
give it a target directory,
or file and it will load it.

42
00:02:54,590 --> 00:02:59,240
So in this case, my file is gonna
be called people-example.csv.

43
00:02:59,240 --> 00:03:02,900
And here we go.

44
00:03:02,900 --> 00:03:07,349
So it prints a few things saying that
it parsed and it parsed it correctly.

45
00:03:07,349 --> 00:03:08,600
And now we have it.

46
00:03:08,600 --> 00:03:11,080
So let's do some SFrame basics.

47
00:03:12,910 --> 00:03:18,661
So let's start with some #SFrame basics.

48
00:03:20,630 --> 00:03:24,620
So some SFrame basics here,
if you just type sf,

49
00:03:24,620 --> 00:03:28,540
enter, in this case shift+enter,
because when they are IPython notebook

50
00:03:28,540 --> 00:03:32,260
it just shows you the first
few lines of that table.

51
00:03:32,260 --> 00:03:38,413
So, in this case I'm going
to put a comment here,#we

52
00:03:38,413 --> 00:03:47,210
can view first few lines of the table.

53
00:03:48,640 --> 00:03:49,870
So, here we go.

54
00:03:49,870 --> 00:03:51,970
This is a very simple table.

55
00:03:51,970 --> 00:03:54,500
Actually, I'm seeing all the lines.

56
00:03:54,500 --> 00:03:56,320
It says the first name Bob,

57
00:03:56,320 --> 00:04:00,400
last name Smith, who lives in
the United States, has age 24.

58
00:04:00,400 --> 00:04:02,790
And we have this for several people.

59
00:04:02,790 --> 00:04:05,320
By the way, there's two ways to
access the first few lines in file.

60
00:04:05,320 --> 00:04:13,980
You can do, type sf, you can also type
sf.head to show the first few lines.

61
00:04:13,980 --> 00:04:16,810
And again, it's a small data
set we're playing with so far.

62
00:04:16,810 --> 00:04:20,436
But there's another command
called sf.tail that

63
00:04:20,436 --> 00:04:24,160
shows the last few lines of the data set.

64
00:04:24,160 --> 00:04:27,200
Which again, this is a small data set.

65
00:04:27,200 --> 00:04:29,262
So first few lines,
last few lines, all of the lines.

66
00:04:29,262 --> 00:04:32,579
[MUSIC]