1
00:00:00,180 --> 00:00:00,510
Hello.

2
00:00:00,540 --> 00:00:01,230
Welcome back.

3
00:00:01,230 --> 00:00:03,390
This is our neural network again.

4
00:00:03,570 --> 00:00:10,080
The green text here indicates the lower numbers and the orange text indicates the number of units in

5
00:00:10,080 --> 00:00:20,830
each layer to compute the z value of Layer 1 we compute the superscript 1 e course W superscript 1 x

6
00:00:21,070 --> 00:00:22,950
plus b superscript 1.

7
00:00:22,990 --> 00:00:26,370
Just like we can see over here X is a column vector.

8
00:00:26,400 --> 00:00:33,670
Therefore its shape is three by one because w superscript 1 which is the weight for Layer 1 has 5 units

9
00:00:33,940 --> 00:00:44,050
and each unit is receiving 3 input then its shape shall be 5 by 3 superscript 1 is the bias for Layer

10
00:00:44,050 --> 00:00:49,310
1 which has 5 unit the shape of B superscript 1 choppy 5 by 1.

11
00:00:49,540 --> 00:00:56,080
If we solve this equation with these input shapes the resultant shape which is the shape of C superscript

12
00:00:56,080 --> 00:00:58,220
1 shall be 5 by 1.

13
00:00:58,270 --> 00:01:04,110
We can actually generalize the shapes of various matrices of a 1 neural network.

14
00:01:04,210 --> 00:01:11,210
Let's see to find capitals of layer L which shall contain the zeal of all training examples.

15
00:01:11,380 --> 00:01:21,220
We compute capital W layer l capital x plus b like we have in this equation over here D show of capital

16
00:01:21,280 --> 00:01:27,790
X will be the number of features which we denote us and superscript 0 by the number of training examples

17
00:01:27,880 --> 00:01:36,070
which we denotes by M the shape of capital W superscript l shall be the number of units of layer L by

18
00:01:36,070 --> 00:01:43,390
the number of features or input the shape of superscript l shall be the number of units of layer L by

19
00:01:43,390 --> 00:01:51,460
the number of examples and the shape of capital Z superscript l shall be the number of units of layer

20
00:01:51,550 --> 00:01:53,890
l by training examples.

21
00:01:53,890 --> 00:02:01,000
In summary the shape of c of a single training example of layer l written here a small c superscript

22
00:02:01,090 --> 00:02:08,860
l is equal to the number of units of layer l by 1 the shape of the vector rise version of c which is

23
00:02:09,160 --> 00:02:16,270
written as capital C is equal to the number of units of layer L by the number of training examples which

24
00:02:16,330 --> 00:02:24,640
is denoted here as m the shape of W superscript l is n superscript l by N superscript l minus 1 the

25
00:02:24,640 --> 00:02:33,460
shape of P superscript l is n superscript l by one the shape of D W superscript l is the same as the

26
00:02:33,460 --> 00:02:41,410
shape of W superscript l the shape of d be superscript l is the same as the shape of P superscript l

27
00:02:41,860 --> 00:02:48,950
let s conclude this lecture by talking a bit about parameters and hyper parameters parameters of a one

28
00:02:48,960 --> 00:02:56,860
new note work out the weight and bias is denoted here as W and b the hyper parameters are the parameters

29
00:02:56,860 --> 00:03:01,220
we can change in order to improve the performance of our new network.

30
00:03:01,300 --> 00:03:06,760
These include the landing rate the number of iterations number of hidden layers number of hitting unit

31
00:03:06,850 --> 00:03:08,890
and activation functions.

32
00:03:08,890 --> 00:03:12,900
So this all the risk for this lesson and I shall see you in the next lesson.