0
1
00:00:00,690 --> 00:00:01,360
Hello.
1

2
00:00:01,410 --> 00:00:02,600
Welcome back.
2

3
00:00:02,610 --> 00:00:10,230
So there are two ways of looking at deep learning and its relationship to the larger field of artificial
3

4
00:00:10,230 --> 00:00:18,030
intelligence. Deep learning can be described as a subset of machine learning which uses artificial neural
4

5
00:00:18,030 --> 00:00:19,840
networks.
5

6
00:00:19,860 --> 00:00:28,560
Machine learning is the study and development of machines that can learn from data. Now because not all
6

7
00:00:28,560 --> 00:00:36,540
deep learning and even not all machine learning is focused around the pursuit of generalized artificial
7

8
00:00:36,540 --> 00:00:37,700
intelligence,
8

9
00:00:38,010 --> 00:00:45,690
We sometimes find the relationship between deep learning and artificial intelligence described as shown
9

10
00:00:45,690 --> 00:00:54,930
in grouping one over here. Generalized artificial intelligence implies building sentient machines,
10

11
00:00:54,960 --> 00:00:59,620
meaning machines that act in all forms like human beings.
11

12
00:00:59,670 --> 00:01:04,710
The simplified way of looking at the relationship between deep learning and artificial intelligence
12

13
00:01:04,770 --> 00:01:11,490
is what we see over here in grouping two, where deep learning is a subset of machine learning, which is
13

14
00:01:11,490 --> 00:01:14,450
in turn a subset of artificial intelligence
14

15
00:01:17,190 --> 00:01:23,210
Supervised learning is a method of transforming one data set into another data set.
15

16
00:01:23,330 --> 00:01:30,440
For example, if we have a data set called February weather and the data set contained recorded weather
16

17
00:01:30,440 --> 00:01:38,480
patterns of the month of February for the last let's say 1000 years and another dataset called April
17

18
00:01:38,480 --> 00:01:45,020
weather and this one contains the weather patterns of the month of April over the same time period,
18

19
00:01:45,730 --> 00:01:52,290
a supervised learning algorithm might try to use one to predict the other.
19

20
00:01:52,700 --> 00:01:59,510
If we are able to successfully train the supervised learning algorithm on the thousand years of February
20

21
00:01:59,510 --> 00:02:07,700
data, of February and April weather data, we would be able to predict
21

22
00:02:07,730 --> 00:02:13,460
the weather of April in the future given the weather of February of the same year.
22

23
00:02:16,190 --> 00:02:25,550
Unsupervised learning finds patterns within a data set. Clustering the data set
23

24
00:02:25,650 --> 00:02:26,990
into groups
24

25
00:02:26,990 --> 00:02:35,090
is a type of unsupervised learning. Clustering transforms a sequence of data points into a sequence
25

26
00:02:35,090 --> 00:02:37,150
of cluster labels.
26

27
00:02:37,310 --> 00:02:44,750
If the unsupervised learning algorithm learns three clusters it is common for these labels to be numbered
27

28
00:02:44,750 --> 00:02:53,490
1 2 and 3. Each data point will be assigned to a number based on which cluster it is in.
28

29
00:02:53,690 --> 00:02:57,470
You can think of unsupervised learning simply as clustering.