1
00:00:00,870 --> 00:00:01,280
Hello.

2
00:00:01,320 --> 00:00:02,550
Welcome back.

3
00:00:02,550 --> 00:00:09,510
Now let's take a look at the multiple input multiple output neural network the multiple inputs multiple

4
00:00:09,570 --> 00:00:17,160
outputs neural network is a combination of the multiple inputs single output in Europe networks.

5
00:00:17,250 --> 00:00:21,810
Now let's say given the temperature humidity and air quality.

6
00:00:22,140 --> 00:00:28,660
Apart from just wanting to predict what a person is happy or sad we also want to predict whether the

7
00:00:28,710 --> 00:00:34,880
sick or healthy as well as predict whether they are active or inactive.

8
00:00:35,670 --> 00:00:38,860
This is what the neural network would look like.

9
00:00:39,000 --> 00:00:46,740
The simplest way to solve this is to view it as a three separate multiple input single output neural

10
00:00:46,750 --> 00:00:48,360
networks.

11
00:00:48,480 --> 00:00:56,100
First we take the input of temperature humidity and air quality.

12
00:00:56,100 --> 00:01:02,130
Multiply each input with its respective weight and add up the three product.

13
00:01:02,130 --> 00:01:05,710
Then we get the prediction for happy or sad.

14
00:01:05,970 --> 00:01:15,650
We repeat the same to predict sick or healthy and then we repeat it again to get active or inactive.

15
00:01:15,660 --> 00:01:21,350
Note that the inputs for predicting the three outputs are the same.

16
00:01:21,900 --> 00:01:25,450
However the weights are different.

17
00:01:25,650 --> 00:01:30,390
Different sets of weight is used to predict the three different output.

18
00:01:30,390 --> 00:01:37,620
We can arrange the weights as a two dimensional array like this the first row represents the weight

19
00:01:37,650 --> 00:01:40,660
for the prediction for sad or happy.

20
00:01:40,680 --> 00:01:48,040
The second rule represents the weight for the prediction for sick or healthy and the third row represent

21
00:01:48,070 --> 00:01:53,610
the weight for the prediction of active or inactive.

22
00:01:53,610 --> 00:02:00,750
In this arrangement the first column in each row represent wait for the temperature inputs.

23
00:02:00,750 --> 00:02:07,920
The second column represents weight for the humidity input and the third represent weight for the air

24
00:02:07,920 --> 00:02:14,560
quality input to compute the multiple input multiple outputs new network.

25
00:02:14,610 --> 00:02:22,080
We have to perform a vector matrix multiplication like this like this function over here.

26
00:02:22,080 --> 00:02:29,670
The vector is the one dimensional array that contains our inputs and The Matrix is the 2 dimensional

27
00:02:29,670 --> 00:02:38,100
array that contains the weight the shadow code will compute the throughput predictions of our multiple

28
00:02:38,100 --> 00:02:41,520
inputs multiple outputs.

29
00:02:41,520 --> 00:02:44,280
You can post a video to take a look at this.

30
00:02:44,310 --> 00:02:46,680
I'm sure it's very straightforward.

31
00:02:46,680 --> 00:02:49,440
Now let's write some code to demonstrate this.