1
00:00:00,630 --> 00:00:01,440
Hello.

2
00:00:01,500 --> 00:00:02,080
Welcome back.

3
00:00:02,610 --> 00:00:06,760
Let's say we have a truly a new network like this.

4
00:00:07,020 --> 00:00:13,370
We don't often include the input layer when we talk about the layers of shared neural network has a

5
00:00:13,370 --> 00:00:20,460
two layer neural network would look like this the input layer is always denoted as a zero in the next

6
00:00:20,460 --> 00:00:26,630
layer is denoted as a one and the one after that is denoted as a to the hidden layer.

7
00:00:26,700 --> 00:00:34,800
A one has 4 unit D superscript represent the Layer number and the subscript represents the unit number.

8
00:00:35,610 --> 00:00:45,260
Let's see how to compute unit a subscript 1 and superscript 2 1 to complement this we begin by compute

9
00:00:45,260 --> 00:00:56,250
in Z subscript 1 z subscript 1 is computed by weight subscript 1 transpose x plus b subscript 1 superscript

10
00:00:56,280 --> 00:00:56,680
1.

11
00:00:56,730 --> 00:01:06,780
As we can see over here remember we use subscript to denote the unit number and superscript to denote

12
00:01:06,900 --> 00:01:10,000
the Layer number.

13
00:01:09,990 --> 00:01:19,510
Once we have found Z subscript 2 1 we apply our activation function to it to get a subscript 1 superscript

14
00:01:19,510 --> 00:01:20,420
2 1.

15
00:01:20,460 --> 00:01:24,820
This is how we find the output of the first unit.

16
00:01:25,650 --> 00:01:32,070
We find the output of the second third and fourth unit using the same method but we have to use our

17
00:01:32,070 --> 00:01:40,500
respective weight and biases like shown in these equations to compute Zi superscript 1 which means the

18
00:01:40,950 --> 00:01:43,660
Zi output of the entire layer.

19
00:01:43,670 --> 00:01:50,940
Number one we've put all of the weight together in a matrix and all of the inputs in effect so and all

20
00:01:50,940 --> 00:01:59,100
of the bias is in effect all and perform you transpose x plus B on the weight matrix and vectors.

21
00:01:59,100 --> 00:02:08,070
The result would look like this like we see over here we end up with a vector containing the ze result

22
00:02:08,580 --> 00:02:13,730
of the of all the units in the layer 1.

23
00:02:13,770 --> 00:02:21,690
As you can see over here by applying our activation function to this resort we get a new vector containing

24
00:02:21,810 --> 00:02:26,130
all the end result of layer one right.

25
00:02:26,190 --> 00:02:28,840
So you can post the video and take a look at this.

26
00:02:28,980 --> 00:02:33,300
This slide over here it's all been arranged nicely.

27
00:02:33,330 --> 00:02:36,140
If you have any questions you can let me know.

28
00:02:36,150 --> 00:02:42,080
Also take note of the dimensions of the parameters the dimension of the weight matrix.

29
00:02:42,110 --> 00:02:50,160
4 by 3 four rows three columns four rows because Layer 1 has four units and three columns because we

30
00:02:50,160 --> 00:02:56,010
have three inputs entering each unit which will spend more time talking about how to fully understand

31
00:02:56,010 --> 00:02:58,190
the dimensions later.

32
00:02:58,260 --> 00:03:01,530
So this order is for this lesson and I shall see you in the next lesson.