1
00:00:00,530 --> 00:00:01,090
I guess.

2
00:00:01,110 --> 00:00:05,220
And welcome to chapter eight point five where we talk about what one encoding.

3
00:00:05,520 --> 00:00:11,900
So as we saw before we did some transmissions on the extreme and X deciliter that's so image data.

4
00:00:11,970 --> 00:00:15,520
Now what about our label little white green and white test.

5
00:00:15,780 --> 00:00:16,530
Well let's find out.

6
00:00:16,530 --> 00:00:19,790
So let's recap what we did on the image data.

7
00:00:19,860 --> 00:00:23,880
We added a four dimension to go from this to this.

8
00:00:23,880 --> 00:00:32,070
We changed it to flow to two data type and we normalized it between 0 and 1 by dividing by 255.

9
00:00:32,450 --> 00:00:33,960
But what do we do to label that.

10
00:00:33,960 --> 00:00:35,450
Now that's true.

11
00:00:35,940 --> 00:00:39,730
So the label is basically in this form for white train.

12
00:00:39,750 --> 00:00:47,850
It's going to be a matrix that is 60000 that has 60000 elements and each element indicates a class label.

13
00:00:47,890 --> 00:00:57,360
So for x for y treatment which is a 28 if this element in white and extreme is going to be 28 by 28

14
00:00:57,840 --> 00:01:00,900
and this zero corresponds to its label here.

15
00:01:01,150 --> 00:01:04,710
However Harris does not use label data like this.

16
00:01:04,890 --> 00:01:06,820
It needs it to be a hot one encoded.

17
00:01:06,900 --> 00:01:09,850
And what does that look like that looks like this.

18
00:01:10,080 --> 00:01:11,640
So we have labels here.

19
00:01:12,030 --> 00:01:20,840
And instead of having a for being represented here it's basically a matrix to has 10 columns now.

20
00:01:21,270 --> 00:01:24,900
So instead of having no call them so effectively want to call them.

21
00:01:24,930 --> 00:01:26,890
Sorry sixty dozen columns.

22
00:01:27,000 --> 00:01:34,580
It has 10 columns here and 60000 rows and each column is a row.

23
00:01:34,590 --> 00:01:39,390
Sorry has basically a 1 0 0 indicating which label it is.

24
00:01:39,690 --> 00:01:42,900
So imagine we have this being transformed.

25
00:01:42,900 --> 00:01:44,090
Sorry let's look at this.

26
00:01:44,090 --> 00:01:45,520
This is a td rule here.

27
00:01:45,810 --> 00:01:48,340
Being transformed into this.

28
00:01:48,370 --> 00:01:55,520
So instead of having this rule before what one coding makes it into this I hope you understand clearly

29
00:01:55,530 --> 00:01:56,600
so we're going to do this now.

30
00:01:56,640 --> 00:01:58,770
You know I write in my book.

31
00:01:59,650 --> 00:01:59,960
OK.

32
00:01:59,970 --> 00:02:06,310
So Step three is a hot one including a full y labels and to do this we basically use any utilities that

33
00:02:06,310 --> 00:02:10,280
are imported from the utilities and all this stuff here.

34
00:02:10,290 --> 00:02:12,600
It just sends it to categorical.

35
00:02:12,600 --> 00:02:17,210
That is how Cara's calls dysfunction Hotpoint including two categorical.

36
00:02:17,460 --> 00:02:18,550
So we have Whitopia.

37
00:02:18,620 --> 00:02:23,160
It's equal to utilities not to categorical and just put the wager in here.

38
00:02:23,310 --> 00:02:25,140
And that transforms it.

39
00:02:25,140 --> 00:02:27,220
So let's take a look at how this actually looks.

40
00:02:27,220 --> 00:02:28,510
So let's run this here.

41
00:02:28,820 --> 00:02:30,120
So no why train.

42
00:02:30,270 --> 00:02:32,420
Let's look at the first rule in waitron.

43
00:02:32,830 --> 00:02:40,700
It's this and basically can see one two three four five six seven eight nine ten elements.

44
00:02:40,920 --> 00:02:45,270
And with this one this looks like the nine and fifth element here.

45
00:02:45,300 --> 00:02:46,920
So this is number five.

46
00:02:46,920 --> 00:02:51,720
So the first element in overtreating data is number five.

47
00:02:52,110 --> 00:02:54,710
So now let's move on to creating a model.