1
00:00:00,100 --> 00:00:00,630
Extreme.

2
00:00:02,230 --> 00:00:06,850
Now, let's start building our artificial neural network model using Get US.

3
00:00:09,320 --> 00:00:16,520
Kiraz is a powerful and easy to use free open source Biton Library for developing and evaluating deep

4
00:00:16,520 --> 00:00:17,330
learning models.

5
00:00:18,730 --> 00:00:20,570
So here we are building artificial.

6
00:00:23,050 --> 00:00:30,640
So artificial neural networks are inspired by biological neural network that constituting brains subsystem

7
00:00:30,640 --> 00:00:38,260
learn to perform tasks, but considering examples generally without being programmed with tasks, specific

8
00:00:38,260 --> 00:00:45,100
roles, rules and is based on a collection of connected units are nodes called artificial neurons,

9
00:00:45,400 --> 00:00:48,700
which loosely model the neuron in a biological brain.

10
00:00:49,420 --> 00:00:51,070
Now let's import Kiraz.

11
00:00:53,640 --> 00:00:54,330
And then.

12
00:00:55,820 --> 00:00:59,030
From general models, import sequential.

13
00:01:00,870 --> 00:01:08,570
Sequential specifies to us that we are creating model sequential and output of each year will add an

14
00:01:08,730 --> 00:01:09,900
input to the next layer.

15
00:01:10,970 --> 00:01:19,510
And then from Kiraz, Dr. Layer's importance, the dances used to specify the full cuddlier classifier

16
00:01:19,520 --> 00:01:20,510
is equal to sequential.

17
00:01:20,660 --> 00:01:23,800
So here we are calling our sequential lurcher, Anderson.

18
00:01:27,560 --> 00:01:36,740
Now, let's add the lawyers in our neural network to our facilities, classified ad dense units is equal

19
00:01:36,740 --> 00:01:37,220
to 10.

20
00:01:39,150 --> 00:01:45,600
Colonel Initialised is uniform activation function is Relu and our input, the mission is 21.

21
00:01:46,590 --> 00:01:52,020
So this is our input important, the output of this layer will be taken as input for this next year,

22
00:01:52,860 --> 00:01:58,920
but she'll have to check all the parameters which are used in this layer.

23
00:02:03,310 --> 00:02:08,470
So here we have units, activation units, Eusebius, Carnel initialised.

24
00:02:11,500 --> 00:02:17,740
So and there is also explanation about that barometer's, so if you want to change any default parameter,

25
00:02:18,130 --> 00:02:20,440
then you can see this document.

26
00:02:21,800 --> 00:02:23,150
And make any changes.

27
00:02:25,540 --> 00:02:27,370
So now we're adding a second layer.

28
00:02:29,060 --> 00:02:36,410
And secondly, we have Colonel initialisms, Unicomp and Activision function is Relu, and here we have

29
00:02:36,410 --> 00:02:42,470
not mentioned input because it has the same input demolition and then we're initializing the output

30
00:02:42,470 --> 00:02:42,820
layer.

31
00:02:43,550 --> 00:02:48,850
So we are classified ad dance units is equal to one colonel.

32
00:02:48,860 --> 00:02:52,660
Initialise that is uniform and activation function is sigmoid.

33
00:02:54,130 --> 00:03:01,270
Now we need to compile our model, so you're classified at compile optimizer is Adam.

34
00:03:02,300 --> 00:03:06,860
Lost is called a binary cross entropy, and Matrix is called accuracy.

35
00:03:08,780 --> 00:03:11,930
Now, later, Anderson, Al Sharpton told Anderson.

36
00:03:15,180 --> 00:03:16,530
Just ignore these warnings.

37
00:03:18,640 --> 00:03:21,840
Now they need to fix our neural network trainings like.

38
00:03:23,380 --> 00:03:28,120
So classify it not fit because other models have been classified in.

39
00:03:29,120 --> 00:03:36,450
So we are putting it on our extreme and vedran and backspaces 10 and number of boxes, hundred.

40
00:03:37,070 --> 00:03:38,840
So we are taking a hundred.

41
00:03:39,350 --> 00:03:45,370
You could take 200 or so, you can increase the boxes and you can increase number of a box available,

42
00:03:45,410 --> 00:03:45,970
Anderson.

43
00:03:46,400 --> 00:03:47,430
And plus you can do.

44
00:03:48,680 --> 00:03:50,750
So if we start training on modern.

45
00:03:58,840 --> 00:04:01,990
So we have successfully presented our model.

46
00:04:03,670 --> 00:04:05,350
Let's check that crazy.

47
00:04:06,590 --> 00:04:09,510
Two zero one six three zero point six three.

48
00:04:10,010 --> 00:04:12,200
So it's nearly constant.

49
00:04:16,640 --> 00:04:20,650
Let's check the final accuracy of the Depok.

50
00:04:22,890 --> 00:04:28,590
Finally, we got accuracy of zero point sixty four point sixty four percent accuracy and loss at zero

51
00:04:28,590 --> 00:04:29,340
point six one.

52
00:04:32,360 --> 00:04:34,410
No, oh, no.

53
00:04:34,490 --> 00:04:35,080
That's just it.

54
00:04:35,630 --> 00:04:40,700
So you're what is going to classify dog predicator function?

55
00:04:42,370 --> 00:04:46,660
And are projecting on our best and we're seven little.

56
00:04:48,380 --> 00:04:49,190
Veteran little.

57
00:04:51,590 --> 00:04:58,330
So here we are taking a conditioner's bread, which is above zero point zero five and started in my

58
00:04:58,790 --> 00:05:00,270
bread, which is a prediction.

59
00:05:01,460 --> 00:05:07,090
So now let's check the accuracy score with an all white predicted and lightest.

60
00:05:08,670 --> 00:05:12,320
So here we got an accuracy of 61 percent on our test.

61
00:05:13,110 --> 00:05:16,710
So we got 64 percent.