1
00:00:00,450 --> 00:00:08,820
If you were looking to generate additional options, an example code, then you must create an advance,

2
00:00:08,820 --> 00:00:09,280
a script.

3
00:00:09,960 --> 00:00:13,590
Let's click here and see what would happen here.

4
00:00:13,610 --> 00:00:15,130
I have my simple a screen.

5
00:00:15,360 --> 00:00:16,960
We already explored that.

6
00:00:17,280 --> 00:00:27,510
Now let's go here and first give it a name and just control s and save your file this time.

7
00:00:27,540 --> 00:00:31,680
This is my advanced.

8
00:00:34,470 --> 00:00:41,170
A script one, and then we can see it here, Atlanta script.

9
00:00:41,820 --> 00:00:48,000
So just like previous one, just like as simple as screen, we need to define our data and don't forget

10
00:00:48,000 --> 00:00:51,320
to add ACLC clear ampoules.

11
00:00:51,330 --> 00:00:57,990
Also, I'm going to copy these codes and just pasted at the beginning of your code.

12
00:00:59,760 --> 00:01:06,090
Here we just generate the data and then this part is the same with the simple cream, if you haven't

13
00:01:06,090 --> 00:01:12,510
watched the video for simplicity, please go back for additional information and the train function,

14
00:01:12,540 --> 00:01:15,420
his train elim back propagation.

15
00:01:15,420 --> 00:01:22,600
And again, you can change it if you want a number of hidden layers generating and that function.

16
00:01:23,250 --> 00:01:24,490
And let's check it here.

17
00:01:25,260 --> 00:01:28,490
This is an important we didn't have it in a simple scrape.

18
00:01:29,610 --> 00:01:39,720
Let's see, what is it giving us notes that input, that process function equals to remove constant

19
00:01:39,720 --> 00:01:40,230
rows.

20
00:01:41,130 --> 00:01:44,400
But what does it mean, remove constant rows?

21
00:01:44,850 --> 00:01:47,880
Let's check our data here in a command box.

22
00:01:47,910 --> 00:01:57,680
I'm going to type S.O.S. then I want to see my data X and Y, X and Y.

23
00:01:57,720 --> 00:02:00,570
Let's just check them here.

24
00:02:01,020 --> 00:02:08,040
We can see some columns and we can see some rows in Matlab find always a sample.

25
00:02:08,040 --> 00:02:14,430
Must be in a column, meaning each row is for and a specific field.

26
00:02:15,180 --> 00:02:24,810
If we have the same data for one field, it means this is a constant data and then it would remove it.

27
00:02:25,330 --> 00:02:32,730
This code will actually tell us go to the net function from the inputs and apply this process function,

28
00:02:32,850 --> 00:02:35,100
remove constant rows.

29
00:02:35,550 --> 00:02:37,110
Let me give you an example.

30
00:02:37,590 --> 00:02:38,420
Here is a Google.

31
00:02:38,430 --> 00:02:46,590
I had this I just conducted a statistical research about the amount of stress level of graduating students.

32
00:02:46,590 --> 00:02:49,710
I gave them a questionnaire with several question.

33
00:02:49,710 --> 00:02:54,390
One of these questions was, what is your year of study?

34
00:02:54,540 --> 00:03:00,840
Of course, they were graduating as students, so they all marked fifth year as students and CS.

35
00:03:00,870 --> 00:03:06,390
All this field in my Excel fine in my data was fifth year.

36
00:03:06,960 --> 00:03:13,590
It actually won't change anything for the output, meaning it doesn't have any affect or it doesn't

37
00:03:13,920 --> 00:03:16,110
change anything for my output.

38
00:03:16,230 --> 00:03:21,750
I can just simply remove it from feeding it to over neural network.

39
00:03:22,150 --> 00:03:28,860
Another example is I have a class which they were 16 students and all of them were female.

40
00:03:28,980 --> 00:03:37,440
Again, I had no choice but to remove all those feel the field of gender here for this specific class

41
00:03:37,440 --> 00:03:40,920
because they were all female, the output wouldn't be any different.

42
00:03:41,400 --> 00:03:44,240
And it's basically the same for this one.

43
00:03:45,300 --> 00:03:46,400
Now, let's check this one.

44
00:03:46,410 --> 00:03:47,610
This is a map.

45
00:03:47,610 --> 00:03:58,110
I mean, Max, this function scales input and target so that they fall in the range of negative one

46
00:03:58,110 --> 00:03:59,040
and one.

47
00:03:59,460 --> 00:04:03,500
This is what we need to do for normalizing the data.

48
00:04:04,080 --> 00:04:11,610
It just says that apply in mapping on a day to also all the output would be between negative one to

49
00:04:11,610 --> 00:04:11,940
one.

50
00:04:13,020 --> 00:04:18,660
This is for pre processing of inputs and this is a post processing for outputs.

51
00:04:18,660 --> 00:04:19,900
It's just doing the same.

52
00:04:20,400 --> 00:04:22,050
Now let's check it here next.

53
00:04:22,050 --> 00:04:24,240
Dot divide function.

54
00:04:24,540 --> 00:04:27,300
Our data would be divided randomly.

55
00:04:27,600 --> 00:04:31,980
And then another field, another function that we have here is divide.

56
00:04:32,020 --> 00:04:37,170
Moad Divide for this example is sample.

57
00:04:37,860 --> 00:04:40,350
Sample is for a static network.

58
00:04:41,010 --> 00:04:45,560
But if you have a dynamic network, then you need to choose time.

59
00:04:45,990 --> 00:04:54,480
So let me just add some explanation as a comment so you can review this for later and you can choose

60
00:04:55,200 --> 00:04:58,440
a different moment based on your data.

61
00:04:59,610 --> 00:05:07,200
It can be time if you have dynamic network.

62
00:05:10,270 --> 00:05:14,110
The other more that we have is sample time.

63
00:05:15,980 --> 00:05:20,540
Sample time, if I'm not wrong, this is also small.

64
00:05:20,930 --> 00:05:22,850
This is to divide.

65
00:05:25,290 --> 00:05:33,060
Targets by both sample and timestep.

66
00:05:36,690 --> 00:05:44,730
The other rule that we have here is all all just means to divide.

67
00:05:49,320 --> 00:05:53,820
To divide up targets by every Eskild value.

68
00:05:56,100 --> 00:06:10,080
OK, and the last mode that we have here is non non means not divide, not divide the time at all,

69
00:06:11,130 --> 00:06:14,190
meaning all the data would be for training.

70
00:06:14,370 --> 00:06:21,540
There won't be any data for validation and testing for it might be useful while you are working with

71
00:06:21,540 --> 00:06:23,880
different data and different samples.

72
00:06:24,030 --> 00:06:29,850
Now, let's just check the next part, which is defined parameters.

73
00:06:30,240 --> 00:06:36,180
70 percent is for training here, 15 percent for validation and 15 percent for testing.

74
00:06:36,510 --> 00:06:40,010
We have this part in the sample script as well.

75
00:06:40,290 --> 00:06:44,470
You can just change them as you need and then let's check it here next.

76
00:06:44,490 --> 00:06:47,700
That perform function is mean square error.

77
00:06:48,030 --> 00:06:54,180
Again, if you need to use another function, you can just go to the help of Matlab and find the name

78
00:06:54,180 --> 00:06:56,280
of different function and add them here.

79
00:06:56,700 --> 00:07:02,050
Next function that we have here is a plot function to see what is a flat function.

80
00:07:02,070 --> 00:07:04,500
Let me show you an example.

81
00:07:05,280 --> 00:07:11,330
I'm going to make this point as a comment by adding a percentage sign.

82
00:07:11,700 --> 00:07:13,980
Go to the editor and click on Ron.

83
00:07:15,910 --> 00:07:22,810
OK, you might already notice the difference here, you don't have actually any plot because I just

84
00:07:22,810 --> 00:07:24,060
removed all of them.

85
00:07:24,520 --> 00:07:34,930
But now let's see what would happen if I had only this board, for example, plot, perform, run again,

86
00:07:35,860 --> 00:07:39,040
and you would see that you have only the performance.

87
00:07:39,040 --> 00:07:45,400
You can see only the performance, but you don't have any other plots that you can just add them.

88
00:07:45,400 --> 00:07:47,890
You can remove them based on your need.

89
00:07:49,720 --> 00:07:52,490
Training of the network will happen in this form.

90
00:07:52,510 --> 00:07:58,270
So before this board, it was just preprocessing and defining the structure.

91
00:07:58,270 --> 00:08:01,900
But we actually didn't train the network in this part.

92
00:08:01,900 --> 00:08:03,210
We can train the network.

93
00:08:03,790 --> 00:08:06,400
I explained it in as simple as Klebb.

94
00:08:06,430 --> 00:08:07,840
Now let's just move on.

95
00:08:07,840 --> 00:08:10,360
We have these portion in a simple script as well.

96
00:08:12,160 --> 00:08:21,550
Twin target tea, which is over target times, it means every element, every sample in the target times

97
00:08:21,850 --> 00:08:28,380
TR, which has some information about overtraining and then here train mask one.

98
00:08:28,930 --> 00:08:30,430
But what is a train mask?

99
00:08:30,670 --> 00:08:38,080
Computer science mask means anything that can give us the altitude of zero and one that may show you

100
00:08:38,080 --> 00:08:42,400
with some example here, T or dot train.

101
00:08:43,750 --> 00:08:44,470
Mask.

102
00:08:50,300 --> 00:08:57,330
OK, Inter, we just make it a little bit bigger so you can see the data.

103
00:08:58,160 --> 00:09:02,160
These are all the data which have been used for training.

104
00:09:02,180 --> 00:09:05,840
You can see these data have been used for training.

105
00:09:05,870 --> 00:09:08,840
This one, we didn't use it for training.

106
00:09:08,840 --> 00:09:11,560
Actually, the neural network didn't use it for the training.

107
00:09:11,570 --> 00:09:16,520
That's why it's a and and I am here means not.

108
00:09:16,520 --> 00:09:19,220
And no, this is not a no.

109
00:09:19,220 --> 00:09:20,090
It hasn't been used.

110
00:09:20,690 --> 00:09:27,260
Of course, it couldn't just put zero because zero is also a number, but it means that we didn't use

111
00:09:27,260 --> 00:09:30,620
these samples for the training process.

112
00:09:30,980 --> 00:09:39,170
If I just click it again and this time instead of train, I just type test, then you should see the

113
00:09:39,170 --> 00:09:39,770
test.

114
00:09:40,010 --> 00:09:44,570
These data, these samples have been used for testing port.

115
00:09:44,810 --> 00:09:46,460
Now let's do something else.

116
00:09:46,470 --> 00:09:54,340
This time I'm going to multiply it by ti our targets and see what would be the output.

117
00:09:54,830 --> 00:10:02,600
Here are the outputs and the targets, outputs, DNA samples and we can see this one, here is one.

118
00:10:02,610 --> 00:10:04,790
So this is giving us an output.

119
00:10:04,910 --> 00:10:19,130
The other thing that we can do here is easier than ti on that test mask and one OK run spot.

120
00:10:19,280 --> 00:10:22,300
Here are all the samples which are none.

121
00:10:22,460 --> 00:10:31,670
But what if we want to see the training only just add and not here not is then this good or let me just

122
00:10:32,060 --> 00:10:35,330
call it here and put it here.

123
00:10:37,080 --> 00:10:38,190
Reverse or not.

124
00:10:38,910 --> 00:10:42,840
These are the items which has been used for training.

125
00:10:43,050 --> 00:10:45,930
The other thing that I can add here is a find.

126
00:10:48,470 --> 00:10:54,870
All not is not test mask data.

127
00:10:56,330 --> 00:10:59,210
In summary, we can see this is number three.

128
00:10:59,780 --> 00:11:04,790
This one is seven three, four, five, six, seven seven, number seven.

129
00:11:05,120 --> 00:11:06,850
And then here is 11.

130
00:11:07,310 --> 00:11:08,340
How many are they?

131
00:11:08,660 --> 00:11:09,860
They are six.

132
00:11:10,670 --> 00:11:13,070
And that's correct, because we have six samples.

133
00:11:13,460 --> 00:11:21,190
If I type here terrorist index, I can save them in a variable.

134
00:11:21,620 --> 00:11:23,270
I just call it a test.

135
00:11:26,730 --> 00:11:31,590
Equals to this can't find just copy this smart.

136
00:11:33,600 --> 00:11:41,730
OK, now it has a name and I just created a variable and I sent all this data that has been selected

137
00:11:41,730 --> 00:11:45,090
for over testing part into this variable.

138
00:11:46,680 --> 00:11:51,620
I can check it all the different training and relevation, let's do it before going to the next part.

139
00:11:51,660 --> 00:11:58,560
If I change this one to train, then I will see the training data.

140
00:11:59,280 --> 00:12:07,980
And again, if I change it to while I can see the validation data or the data that has been selected

141
00:12:07,980 --> 00:12:08,910
for the validation.

142
00:12:09,150 --> 00:12:12,610
And since world of testing and validation are 15 percent.

143
00:12:12,630 --> 00:12:20,520
So in the number of samples that we have, both of them are six can just use these math to separate

144
00:12:20,790 --> 00:12:27,090
the samples and see which one is for test and which one is for training processes.

145
00:12:27,780 --> 00:12:32,700
Now, let's back there in our Advanta scrape.

146
00:12:33,150 --> 00:12:36,360
The other data that we have here are the plotting.

147
00:12:36,540 --> 00:12:44,640
We had this sport in a simple script as well, and that's basically all in the next session, I'm going

148
00:12:44,640 --> 00:12:49,860
to create another advance, a script, and we will try to change the structure.