en_Supervised vs Unsupervised.srt

0
00:00:00,539 --> 00:00:02,460
Hello, and welcome!

1
00:00:02,460 --> 00:00:08,389
In this video, we’ll introduce supervised algorithms versus unsupervised algorithms.

2
00:00:08,389 --> 00:00:10,690
So let’s get started.

3
00:00:10,690 --> 00:00:15,810
An easy way to begin grasping the concept of supervised learning is by looking directly

4
00:00:15,810 --> 00:00:19,060
at the words that make it up.

5
00:00:19,060 --> 00:00:24,870
Supervise means to observe and direct the execution of a task, project, or activity.

6
00:00:24,870 --> 00:00:28,330
Obviously, we aren’t going to be supervising a person…

7
00:00:28,330 --> 00:00:33,730
Instead, we’ll be supervising a machine learning model that might be able to produce

8
00:00:33,730 --> 00:00:37,020
classification regions like we see here.

9
00:00:37,020 --> 00:00:42,090
So, how do we supervise a machine learning model?

10
00:00:42,090 --> 00:00:44,360
We do this by “teaching” the model.

11
00:00:44,360 --> 00:00:50,400
That is, we load the model with knowledge so that we can have it predict future instances.

12
00:00:50,400 --> 00:00:56,810
But … this leads to the next question, which is, “How exactly do we teach a model?”

13
00:00:56,810 --> 00:01:01,260
We teach the model by training it with some data from a labeled dataset.

14
00:01:01,260 --> 00:01:05,370
It’s important to note that the data is labeled.

15
00:01:05,370 --> 00:01:07,660
And what does a labeled dataset look like?

16
00:01:07,660 --> 00:01:10,440
Well, it can look something like this.

17
00:01:10,440 --> 00:01:14,630
This example is taken from the cancer dataset.

18
00:01:14,630 --> 00:01:19,810
As you can see, we have some historical data for patients, and we already know the class

19
00:01:19,810 --> 00:01:21,160
of each row.

20
00:01:21,160 --> 00:01:26,000
Let’s start by introducing some components of this table.

21
00:01:26,000 --> 00:01:31,970
The names up here, which are called Clump thickness, Uniformity of cell size, Uniformity

22
00:01:31,970 --> 00:01:38,570
of cell shape, Marginal adhesion, and so on, are called Attributes.

23
00:01:38,570 --> 00:01:43,230
The columns are called Features, which include the data.

24
00:01:43,230 --> 00:01:48,380
If you plot this data, and look at a single data point on a plot, it’ll have all of

25
00:01:48,380 --> 00:01:50,299
these attributes.

26
00:01:50,299 --> 00:01:55,280
That would make a row on this chart, also referred to as an observation.

27
00:01:55,280 --> 00:02:01,030
Looking directly at the value of the data, you can have two kinds.

28
00:02:01,030 --> 00:02:03,280
The first is numerical.

29
00:02:03,280 --> 00:02:08,530
When dealing with machine learning, the most commonly used data is numeric.

30
00:02:08,530 --> 00:02:14,090
The second is categorical… that is, it’s non-numeric, because it contains characters

31
00:02:14,090 --> 00:02:15,319
rather than numbers.

32
00:02:15,319 --> 00:02:22,400
In this case, it’s categorical because this dataset is made for Classification.

33
00:02:22,400 --> 00:02:25,730
There are two types of Supervised Learning techniques.

34
00:02:25,730 --> 00:02:30,989
They are: classification and regression.

35
00:02:30,989 --> 00:02:37,480
Classification is the process of predicting a discrete class label or category.

36
00:02:37,480 --> 00:02:41,980
Regression is the process of predicting a continuous value as opposed to predicting

37
00:02:41,980 --> 00:02:45,900
a categorical value in Classification.

38
00:02:45,900 --> 00:02:47,700
Look at this dataset.

39
00:02:47,700 --> 00:02:51,489
It is related to Co2 emissions of different cars.

40
00:02:51,489 --> 00:02:59,360
It includes Engine size, Cylinders, Fuel Consumption and Co2 emission of various models of automobiles.

41
00:02:59,360 --> 00:03:05,780
Given this dataset, you can use regression to predict the Co2 emission of a new car by

42
00:03:05,780 --> 00:03:11,620
using other fields, such as Engine size or number of Cylinders.

43
00:03:11,620 --> 00:03:14,340
  Since we know the meaning of supervised learning,

44
00:03:14,340 --> 00:03:17,900
what do you think unsupervised learning means?

45
00:03:17,900 --> 00:03:19,400
Yes!

46
00:03:19,400 --> 00:03:22,099
Unsupervised Learning is exactly as it sounds.

47
00:03:22,099 --> 00:03:27,730
We do not supervise the model, but we let the model work on its own to discover information

48
00:03:27,730 --> 00:03:31,140
that may not be visible to the human eye.

49
00:03:31,140 --> 00:03:39,000
It means, The Unsupervised algorithm trains on the dataset, and draws conclusions on UNLABELED

50
00:03:39,000 --> 00:03:40,249
data.

51
00:03:40,249 --> 00:03:45,519
Generally speaking, unsupervised learning has more difficult algorithms than supervised

52
00:03:45,519 --> 00:03:51,069
learning, since we know little to no information about the data, or the outcomes that are to

53
00:03:51,069 --> 00:03:52,409
be expected.

54
00:03:52,409 --> 00:03:59,939
Dimension reduction, Density estimation, Market basket analysis and Clustering are the most

55
00:03:59,939 --> 00:04:04,739
widely used unsupervised machine learning techniques.

56
00:04:04,739 --> 00:04:11,229
Dimensionality Reduction and/or feature selection play a large role in this by reducing redundant

57
00:04:11,229 --> 00:04:15,250
features to make the classification easier.

58
00:04:15,250 --> 00:04:20,840
Market basket analysis is a modelling technique based upon the theory that if you buy a certain

59
00:04:20,839 --> 00:04:25,640
group of items, you’re more likely to buy another group of items.

60
00:04:25,640 --> 00:04:31,100
Density estimation is a very simple concept that is mostly used to explore the data to

61
00:04:31,100 --> 00:04:33,610
find some structure within it.

62
00:04:33,610 --> 00:04:36,340
And finally, clustering.

63
00:04:36,340 --> 00:04:40,650
Clustering is considered to be one of the most popular unsupervised machine learning

64
00:04:40,650 --> 00:04:46,770
techniques used for grouping data points or objects that are somehow similar.

65
00:04:46,770 --> 00:04:52,190
Cluster analysis has many applications in different domains, whether it be a bank’s

66
00:04:52,190 --> 00:04:58,070
desire to segment its customers based on certain characteristics, or helping an individual

67
00:04:58,070 --> 00:05:02,640
to organize and group his/her favourite types of music!

68
00:05:02,640 --> 00:05:09,080
Generally speaking, though, Clustering is used mostly for: Discovering structure, Summarization,

69
00:05:09,080 --> 00:05:10,950
and Anomaly detection.

70
00:05:10,950 --> 00:05:17,790
So, to recap, the biggest difference between Supervised and Unsupervised Learning is that

71
00:05:17,790 --> 00:05:24,290
supervised learning deals with labeled data while Unsupervised Learning deals with unlabeled

72
00:05:24,290 --> 00:05:25,390
data.

73
00:05:25,390 --> 00:05:31,810
In supervised learning, we have machine learning algorithms for Classification and Regression.

74
00:05:31,810 --> 00:05:36,200
In unsupervised learning, we have methods such as clustering.

75
00:05:36,200 --> 00:05:43,640
In comparison to supervised learning, unsupervised learning has fewer models and fewer evaluation

76
00:05:43,640 --> 00:05:49,290
methods that can be used to ensure that the outcome of the model is accurate.

77
00:05:49,290 --> 00:05:54,840
As such, unsupervised learning creates a less controllable environment, as the machine is

78
00:05:54,840 --> 00:05:57,910
creating outcomes for us.

79
00:05:57,910 --> 00:05:58,800
Thanks for watching!