1
00:00:01,830 --> 00:00:10,260
OK, let's start working the simplest craib, just click on it and then you will have new window here

2
00:00:10,260 --> 00:00:11,170
and your Massola.

3
00:00:11,580 --> 00:00:17,160
This is a simple script and it's giving us some information.

4
00:00:17,640 --> 00:00:20,480
Take a look at some of these informations here.

5
00:00:20,490 --> 00:00:26,910
We have X, which is of our input equals X, the boat input of our system and neural network.

6
00:00:26,940 --> 00:00:30,300
Of course, we are feeding it with the same data.

7
00:00:30,540 --> 00:00:36,560
So there are the same and then teasers then for Target the our target.

8
00:00:36,570 --> 00:00:41,670
We want it to be equal to Y, which is output of our system.

9
00:00:42,180 --> 00:00:50,070
This one is between function and for this particular example, we are using train L.M., which is a

10
00:00:50,070 --> 00:00:53,560
stand for Leverne Berg back propagation.

11
00:00:53,880 --> 00:00:58,110
There are several, I agree, Tim, that you can use for training your function.

12
00:00:58,380 --> 00:01:05,220
If you are looking for any specific algorithm, you can just search in a help box of your matlab and

13
00:01:05,220 --> 00:01:10,620
find a code of Matlab for that algorithm and then copy it here.

14
00:01:11,250 --> 00:01:17,620
Next, information that we have here is number of Naranj and our hidden layer, and that's three.

15
00:01:17,640 --> 00:01:21,710
This is what we said while we were working with the unfitting tool.

16
00:01:21,870 --> 00:01:23,220
You can change it here.

17
00:01:23,400 --> 00:01:27,620
And then the next one is Feed Tonette function.

18
00:01:27,780 --> 00:01:35,670
We are saying that from Freaknik function, get the train function and then number of neurons in each

19
00:01:35,670 --> 00:01:40,570
layer and put them into an object name net this.

20
00:01:40,680 --> 00:01:42,150
It has several features.

21
00:01:42,330 --> 00:01:51,990
One of these features is Divide Parro, which is for dividing the data into the training ratio, validation,

22
00:01:51,990 --> 00:01:53,630
Rafia and test ratio.

23
00:01:54,000 --> 00:02:02,040
And as you can see here, we just didn't change the default, which was 17 15 percent for validation

24
00:02:02,040 --> 00:02:03,920
and 15 percent for testing.

25
00:02:04,290 --> 00:02:07,770
You can change them in the program here as well.

26
00:02:08,100 --> 00:02:10,170
We already design our network.

27
00:02:10,170 --> 00:02:11,700
We have over neural network.

28
00:02:11,910 --> 00:02:14,520
We define the number of neurons.

29
00:02:14,520 --> 00:02:21,540
We defined algorithm to be used, but we have to start training, get from this part.

30
00:02:21,540 --> 00:02:24,210
We will start training our network.

31
00:02:24,660 --> 00:02:26,250
So it's using train.

32
00:02:26,430 --> 00:02:32,900
It gets three inputs net, which is our function of NAV.

33
00:02:32,910 --> 00:02:34,860
This is our object that we created.

34
00:02:35,400 --> 00:02:41,480
And then here we have X inputs and finally why we choose the target outputs.

35
00:02:42,240 --> 00:02:50,260
This one will give us two parameter net, which is our net signal and then another one named TR.

36
00:02:50,520 --> 00:02:54,570
This TR has some information about the training process.

37
00:02:55,350 --> 00:02:58,050
Let me just change you to a space then.

38
00:02:58,410 --> 00:03:01,410
Let's check it here for the test network.

39
00:03:01,410 --> 00:03:10,680
The net of X would be equal to we will send equal here in programming means we will set Y to the net

40
00:03:10,700 --> 00:03:11,580
of X.

41
00:03:11,580 --> 00:03:20,640
It means the data from here goes to our variable Y and we have another variable name E we want this

42
00:03:20,640 --> 00:03:24,570
E to have the G subtract T and Y what.

43
00:03:24,570 --> 00:03:26,910
What is this G subtribe.

44
00:03:26,920 --> 00:03:29,150
Is this then for generalize.

45
00:03:29,160 --> 00:03:38,040
Subtract it, we'll just subtract the two sets of data which have the same format.

46
00:03:38,040 --> 00:03:44,430
For example, here there are just numbers ts for our target and why is the output of our system.

47
00:03:44,700 --> 00:03:52,650
So we want to subtract the output off of our neural network from the output of our system and find out

48
00:03:52,650 --> 00:03:54,750
what is upper error here.

49
00:03:54,750 --> 00:04:03,840
We have another function perform and it gets three variables net target in Y and finally we will send

50
00:04:03,840 --> 00:04:06,420
it to another variable name performance.

51
00:04:07,110 --> 00:04:14,500
This net here, you can call it anywhere in the program and then it will open for you and fifteen to

52
00:04:14,550 --> 00:04:17,490
window before running the program.

53
00:04:17,490 --> 00:04:18,870
We need to do something.

54
00:04:18,870 --> 00:04:22,800
First of all, of course, control is to save your file.

55
00:04:22,800 --> 00:04:27,450
I'm going to complete a simple script.

56
00:04:28,920 --> 00:04:29,420
One.

57
00:04:31,950 --> 00:04:34,990
OK, here we can see over fine here.

58
00:04:35,700 --> 00:04:38,280
Then we need to define our data.

59
00:04:38,300 --> 00:04:39,590
We haven't decided yet.

60
00:04:39,600 --> 00:04:42,180
X is equal to the space.

61
00:04:44,790 --> 00:04:58,320
From zero to two pi, and this time let's try another network with 40 samples and then four, why,

62
00:04:58,320 --> 00:05:01,200
of course this is equal to sign this off X.

63
00:05:01,980 --> 00:05:08,370
The other thing that I'm of recommending to add in a beginning of every program that you have while

64
00:05:08,370 --> 00:05:14,660
working with neural networks First Codicils ACLC to just clear your command window.

65
00:05:15,000 --> 00:05:19,200
The other one is to clear the memory.

66
00:05:20,790 --> 00:05:24,870
If you don't put these in some situation, you might get some error.

67
00:05:26,240 --> 00:05:28,380
Another one is close.

68
00:05:28,650 --> 00:05:32,910
All this one bill pulls all the open graphs.

69
00:05:33,660 --> 00:05:39,390
You can manage your data better now all six.

70
00:05:39,390 --> 00:05:41,190
Let's click on Run.

71
00:05:44,620 --> 00:05:51,450
OK, the first thing that you might notice in that comment window is the performance, the performance.

72
00:05:51,460 --> 00:05:54,430
This is the way she referred to performance.

73
00:05:54,910 --> 00:06:02,860
And here we can see that and fitting tool that we were opening before using the vertical user interface.

74
00:06:02,880 --> 00:06:08,310
This time we just opened up by running the program, which Matlab generated for us.

75
00:06:08,770 --> 00:06:13,290
So it's just giving some information, like in a hidden layer.

76
00:06:13,300 --> 00:06:15,490
We have to inherent in the output layer.

77
00:06:15,520 --> 00:06:19,510
We have one now and we can change number of neurons in the output layer.

78
00:06:20,140 --> 00:06:22,110
Let me close this one here.

79
00:06:22,120 --> 00:06:26,180
You can see that validation check is top overtraining.

80
00:06:26,410 --> 00:06:28,030
Let's see the performance.

81
00:06:29,650 --> 00:06:33,760
OK, and you have some other information, like let's check the feet.

82
00:06:34,560 --> 00:06:36,030
OK, this is a good fit.

83
00:06:36,040 --> 00:06:37,660
It's just like a sinus wave.

84
00:06:37,660 --> 00:06:42,010
And that's what we're looking for from this neural network.

85
00:06:43,270 --> 00:06:45,370
Now, let me close this one.

86
00:06:45,670 --> 00:06:52,330
If you are writing a paper or you want to have some particular report, you probably need different

87
00:06:52,330 --> 00:06:53,020
plotting.

88
00:06:53,860 --> 00:06:55,510
As you might already guess.

89
00:06:55,510 --> 00:07:01,780
This percentage sign will make comment, will make the line as a comment.

90
00:07:02,140 --> 00:07:03,370
If I remove that.

91
00:07:03,490 --> 00:07:06,700
You can see this is now active and it's in the color of black.

92
00:07:06,700 --> 00:07:13,630
So it's active and the active two of them probably we don't need the regression.

93
00:07:13,930 --> 00:07:19,810
However, you might need it for analyzing part of your data so you can just make them comment or uncommon.

94
00:07:19,990 --> 00:07:28,270
But let's see what would happen if I just have the regression and fit this jacket here.

95
00:07:29,060 --> 00:07:33,670
It will first train my network and then it will open to window.

96
00:07:35,650 --> 00:07:43,810
This simple creep has limited features and functions, if you're looking for, modify your network more

97
00:07:43,810 --> 00:07:48,470
and you need to use more features and functions of Matlab.

98
00:07:48,790 --> 00:07:51,160
Then you need to open advanced a script.

99
00:07:51,370 --> 00:07:55,750
In the next session, I'm going to explain advance the script.