1
00:00:00,950 --> 00:00:08,510
Highlight, how about some of the important hubs when come with neural network, deep learning, machine

2
00:00:08,510 --> 00:00:11,560
learning and I so you will meet them.

3
00:00:12,490 --> 00:00:18,620
A lot of time now let's talk about so the first one is Adam.

4
00:00:21,590 --> 00:00:22,490
Optimizer.

5
00:00:25,350 --> 00:00:26,490
So what are they?

6
00:00:31,040 --> 00:00:31,760
A un.

7
00:00:33,860 --> 00:00:34,450
Will.

8
00:00:36,890 --> 00:00:40,400
For for first order.

9
00:00:44,920 --> 00:00:45,220
And.

10
00:00:47,450 --> 00:00:47,930
By.

11
00:00:48,980 --> 00:00:52,000
Optimization are.

12
00:00:55,800 --> 00:00:57,000
Stochastic.

13
00:01:01,040 --> 00:01:02,360
Objective function.

14
00:01:04,580 --> 00:01:05,990
Based on.

15
00:01:09,360 --> 00:01:09,930
Dr..

16
00:01:12,680 --> 00:01:17,960
Estimates are lower order.

17
00:01:19,100 --> 00:01:19,760
Momani.

18
00:01:22,490 --> 00:01:30,620
The next one I want to talk about is men underscore to underscore value.

19
00:01:31,650 --> 00:01:32,460
Error, I mean.

20
00:01:35,010 --> 00:01:39,390
And this a lost function.

21
00:01:41,610 --> 00:01:42,450
Sharmeen.

22
00:01:44,510 --> 00:01:45,260
Square.

23
00:01:47,440 --> 00:01:53,170
Error usually, right, shot at him as a.

24
00:01:56,950 --> 00:01:57,850
Majles.

25
00:01:59,830 --> 00:02:01,030
The average.

26
00:02:02,930 --> 00:02:05,990
The squares is.

27
00:02:09,020 --> 00:02:11,300
How the arrow is.

28
00:02:15,310 --> 00:02:21,850
That is the average square difference.

29
00:02:23,560 --> 00:02:24,310
pitchIN.

30
00:02:25,900 --> 00:02:26,590
The.

31
00:02:29,740 --> 00:02:39,550
Let me make a new life for the show a little bit long, so between the.

32
00:02:41,800 --> 00:02:42,670
Actual.

33
00:02:46,300 --> 00:02:48,760
Estimated value.

34
00:02:50,320 --> 00:02:55,280
And what's to be estimated?

35
00:02:57,220 --> 00:03:04,330
MSA, a measure of the a quality.

36
00:03:07,480 --> 00:03:09,010
An estimated.

37
00:03:14,440 --> 00:03:17,290
It's always.

38
00:03:24,270 --> 00:03:25,380
Non-negative.

39
00:03:28,170 --> 00:03:29,080
And, uh.

40
00:03:30,330 --> 00:03:31,170
Clothes are.

41
00:03:32,810 --> 00:03:35,930
The value add to zero.

42
00:03:38,550 --> 00:03:42,910
They are better, so, of course, we want our role to be zero.

43
00:03:43,860 --> 00:03:49,230
So out of it, we have a big our model is going to be very worse.

44
00:03:50,190 --> 00:03:57,510
And the next one I want to talk about is I could raise a metric.

45
00:03:59,040 --> 00:04:00,210
So metric.

46
00:04:01,730 --> 00:04:15,440
Is our focus on that issue to evaluate the performance of the war or the.

47
00:04:16,680 --> 00:04:18,240
Training and testing.

48
00:04:20,400 --> 00:04:23,430
Now, let's talk about more.

49
00:04:26,350 --> 00:04:31,000
On the outer perimeter, we will saw.

50
00:04:33,070 --> 00:04:34,900
Fred, underscore train.

51
00:04:36,490 --> 00:04:40,060
So this is on a real.

52
00:04:44,140 --> 00:04:45,160
Predictors.

53
00:04:48,830 --> 00:04:49,850
Training data.

54
00:04:53,870 --> 00:05:02,690
RESPA underscore a train, so this a un arri.

55
00:05:08,260 --> 00:05:09,420
Responded to.

56
00:05:10,720 --> 00:05:11,350
Epoch.

57
00:05:15,500 --> 00:05:16,680
Equal 1000.

58
00:05:19,070 --> 00:05:24,410
This is the number of epochs, the.

59
00:05:26,020 --> 00:05:38,410
Trained model and EPOP is an iteration over the entire.

60
00:05:39,560 --> 00:05:41,840
X and Y.

61
00:05:43,280 --> 00:05:44,390
Data provided.

62
00:05:46,920 --> 00:05:47,550
And.

63
00:05:48,860 --> 00:05:49,700
The Post.

64
00:05:52,040 --> 00:06:00,920
Equal one, please, God, please, not that you need to keep this, but while we train the model.

65
00:06:02,680 --> 00:06:08,280
So this is an in danger.

66
00:06:10,890 --> 00:06:13,560
So zero, one and two.

67
00:06:18,920 --> 00:06:20,270
Here, the.

68
00:06:23,400 --> 00:06:24,210
He earns.

69
00:06:27,350 --> 00:06:34,310
For shirt, they might represent the.

70
00:06:35,360 --> 00:06:36,110
Following.

71
00:06:38,260 --> 00:06:52,810
Zero is silane, one ACOA Progress Bar and two equal, one lie per epoch.

72
00:06:57,080 --> 00:07:00,770
So far, worse, boss.

73
00:07:02,950 --> 00:07:04,150
Wise forward, boss.

74
00:07:07,140 --> 00:07:07,740
So.

75
00:07:11,200 --> 00:07:12,550
Follow us by.

76
00:07:14,870 --> 00:07:15,740
Implies.

77
00:07:18,500 --> 00:07:21,020
For propagating.

78
00:07:25,830 --> 00:07:26,720
From the.

79
00:07:28,500 --> 00:07:29,220
In July.

80
00:07:31,700 --> 00:07:32,420
Today.

81
00:07:34,140 --> 00:07:35,180
I would like to.

82
00:07:39,620 --> 00:07:40,280
Backwords.

83
00:07:43,160 --> 00:07:43,640
But.

84
00:07:45,620 --> 00:07:46,550
A.

85
00:07:51,060 --> 00:07:52,300
Might work, but.

86
00:07:54,580 --> 00:07:55,080
Ms.

87
00:07:55,480 --> 00:07:56,110
Bligh's.

88
00:08:03,170 --> 00:08:03,480
Propp.

89
00:08:05,520 --> 00:08:06,420
Well, getting.

90
00:08:08,270 --> 00:08:09,050
So.

91
00:08:12,390 --> 00:08:13,470
Back propagating.

92
00:08:17,240 --> 00:08:18,200
From the.

93
00:08:21,730 --> 00:08:25,390
I would like to the in Paul.

94
00:08:29,630 --> 00:08:37,630
They are sort of our it is far more important I would hand back is from the elbow to the elbow.

95
00:08:42,270 --> 00:08:43,860
So let.

96
00:08:46,360 --> 00:08:47,490
Put a dot in here.

97
00:08:55,410 --> 00:08:59,880
And now that horrible number of iteration.

98
00:09:01,260 --> 00:09:04,110
So no iteration.

99
00:09:06,760 --> 00:09:07,270
A.

100
00:09:09,220 --> 00:09:12,100
There are a number of iteration.

101
00:09:14,340 --> 00:09:27,510
Implies no number of buses where one bus equals one four bus.

102
00:09:28,480 --> 00:09:31,720
Plus one.

103
00:09:34,620 --> 00:09:46,650
That was Basil Dinamo iteration as the number of passes where one was equal, one for spies, plus one

104
00:09:46,670 --> 00:09:49,740
backwards, Bizo, the total of the forward and was.

105
00:09:53,000 --> 00:10:06,530
And now, for example, now that resembles order, it will have 1200, I say 12000 example, so it will

106
00:10:06,530 --> 00:10:10,370
have 12000 training Sambos.

107
00:10:12,520 --> 00:10:18,400
And that our Boschi is and our.

108
00:10:20,130 --> 00:10:21,510
Boschi is

109
00:10:25,980 --> 00:10:38,880
six thousand, so it will take us to iteration, so it will have two iterations to complete.

110
00:10:41,690 --> 00:10:51,530
To complete one epoch, so that is in the literature that is in the front is the reason we bought the

111
00:10:51,530 --> 00:10:58,190
first six thousand symbols and both from a forward base and a bicycle pass.

112
00:11:00,190 --> 00:11:07,790
In the second iteration, we bought the ninety thousand Sambos and both form of our bus and a bus.

113
00:11:08,200 --> 00:11:16,030
So after two iterations, our neural network will see the how trought how that resembles which it one

114
00:11:16,030 --> 00:11:16,690
a box.

115
00:11:18,510 --> 00:11:24,150
Now, that hobel too positive and too negative, so.

116
00:11:25,380 --> 00:11:27,090
That's right, and thanks for the.

117
00:11:29,520 --> 00:11:30,840
So to.

118
00:11:32,180 --> 00:11:36,860
Positive is an outcome.

119
00:11:39,110 --> 00:11:41,360
Where the model.

120
00:11:42,960 --> 00:11:45,900
Correctly predict.

121
00:11:51,500 --> 00:11:52,150
The.

122
00:11:55,450 --> 00:11:56,130
So.

123
00:11:57,810 --> 00:12:02,880
Let's say a true narrative is an outcome where the model correctly predict.

124
00:12:04,180 --> 00:12:08,380
The negative plus name that he last.

125
00:12:10,070 --> 00:12:21,110
So a false positive is an outcome where the model.

126
00:12:24,370 --> 00:12:24,880
Predict.

127
00:12:26,290 --> 00:12:26,920
The.

128
00:12:36,050 --> 00:12:36,650
Fose.

129
00:12:38,940 --> 00:12:49,050
Positive is an outcome where the model incorrectly, the model predicts incorrectly.

130
00:12:52,160 --> 00:12:52,790
The.

131
00:12:55,750 --> 00:12:56,740
Positive class.

132
00:12:59,530 --> 00:13:01,480
Next one, we were talk about a fall

133
00:13:04,630 --> 00:13:05,350
nagati.

134
00:13:06,730 --> 00:13:07,690
Is an.

135
00:13:10,220 --> 00:13:11,030
I'll come.

136
00:13:17,450 --> 00:13:21,140
The model predicts.

137
00:13:25,050 --> 00:13:25,530
In.

138
00:13:29,230 --> 00:13:35,560
Correct, lay the glass.

139
00:13:37,240 --> 00:13:43,450
So negative for what are they, a false negative and.

140
00:13:44,850 --> 00:13:46,860
The last one will be a true.

141
00:13:47,950 --> 00:13:56,370
Bossidy, so how come where the model correctly predict the positive cost?

142
00:14:01,390 --> 00:14:02,740
So later we will meet.

143
00:14:05,050 --> 00:14:11,730
A lot of positive or negative for a true negative and positive, shall we?

144
00:14:12,490 --> 00:14:15,190
I hope you will take this note.

145
00:14:16,400 --> 00:14:23,780
And you can understand it easier, much easier when we come to do the real world project.

146
00:14:25,010 --> 00:14:32,450
And the last one is how we talk about a force that to train.

147
00:14:33,420 --> 00:14:34,320
Neural network.

148
00:14:35,930 --> 00:14:37,670
So what does that.

149
00:14:39,600 --> 00:14:40,860
The first time a.

150
00:14:46,270 --> 00:14:47,050
Import.

151
00:14:49,050 --> 00:14:54,120
Sequential glass from Carus Dot model.

152
00:14:58,970 --> 00:15:00,620
So the second step is.

153
00:15:02,300 --> 00:15:03,040
Layers.

154
00:15:04,430 --> 00:15:06,230
Using DOT at.

155
00:15:08,380 --> 00:15:09,040
Method.

156
00:15:10,150 --> 00:15:11,920
That's just a.

157
00:15:14,210 --> 00:15:15,410
Can figure.

158
00:15:18,470 --> 00:15:20,330
The London.

159
00:15:21,560 --> 00:15:24,880
Process using not.

160
00:15:30,030 --> 00:15:30,870
Kombai.

161
00:15:32,090 --> 00:15:43,790
Made hard and the last time a grand model on 28, I said you did not fit.

162
00:15:46,160 --> 00:15:55,520
Method, and that is on in this video and this note as well, so this explain everything for you.

163
00:15:55,880 --> 00:16:03,620
And I hope when you build a real model, you can bring out next to you so you can understand it.

164
00:16:03,980 --> 00:16:04,200
And.

165
00:16:05,300 --> 00:16:12,800
So if you take this note during and we train the model on the real goal now, it's much easier to understand

166
00:16:12,800 --> 00:16:15,070
why we do this and what do they mean.

167
00:16:15,620 --> 00:16:17,910
And that is the end of this video.

168
00:16:18,080 --> 00:16:19,190
I hope you enjoy it.

169
00:16:19,490 --> 00:16:21,380
And I will see you in the next video.