1
00:00:00,610 --> 00:00:02,120
Welcome back.

2
00:00:02,140 --> 00:00:08,050
Finally let's examine the neural network with a hidden layer.

3
00:00:08,110 --> 00:00:16,030
The neural network with one hidden layer is this combination of the multiple inputs single output and

4
00:00:16,380 --> 00:00:20,920
the single input multiple output neural network.

5
00:00:20,920 --> 00:00:29,590
Essentially we take the output of one layer and feed it as input of another layer later on in the course.

6
00:00:29,620 --> 00:00:34,620
You understand why our neural network will require hidden layers.

7
00:00:34,660 --> 00:00:38,920
Let's see how the prediction of side can be derived.

8
00:00:38,930 --> 00:00:46,180
We stuff from the inputs by taking the temperature input multiplying it by its correspondent weight

9
00:00:47,030 --> 00:00:54,550
taken the humidity input multiplying it also by its corresponding weight and then taken the air quality

10
00:00:54,640 --> 00:00:58,330
input multiplying it by its correspondent weight.

11
00:00:58,480 --> 00:01:01,560
We then add the three products together.

12
00:01:01,660 --> 00:01:04,660
This will give us the hidden zero.

13
00:01:04,660 --> 00:01:12,880
We then repeat the same sequence to derive it in one and it into after we have derived hidden zero it

14
00:01:12,890 --> 00:01:14,860
in one and hit in two.

15
00:01:14,950 --> 00:01:23,530
We multiply it in zero in one and it in two by the corresponding weight and then we add up to three

16
00:01:23,530 --> 00:01:29,380
product to get the prediction offside to compute all predictions.

17
00:01:29,380 --> 00:01:35,920
We will have to perform two consecutive factor matrix multiplications.

18
00:01:35,920 --> 00:01:43,180
We will have two sets of weight the first set will be the weight for the inputs to hit lay out computation

19
00:01:43,920 --> 00:01:51,940
and the second set will be the weight for the hidden layer to output layer computation to compute our

20
00:01:51,940 --> 00:01:52,720
neural network.

21
00:01:52,720 --> 00:02:02,800
We first multiply our vector of input by the matrix of input to hidden weight.

22
00:02:02,800 --> 00:02:10,840
This should result in a vector and like I mentioned earlier a vector is simply a one dimensional array

23
00:02:11,410 --> 00:02:15,900
this vector show contain hidden zero hidden one on it and.

24
00:02:16,110 --> 00:02:24,510
We take this vector and multiplied by the hit into output weight matrix this vector matrix multiplications

25
00:02:24,570 --> 00:02:32,580
shall also result in effect for the vector shall contain the predictions for sad sake and active a simple

26
00:02:32,580 --> 00:02:33,060
cluster.

27
00:02:33,330 --> 00:02:36,420
Let's write some code to demonstrate what we've just broken ball.