1
00:00:01,830 --> 00:00:08,490
Of implementation up and error in this session, we will see how can we implement a simple neuron using

2
00:00:08,490 --> 00:00:08,980
hardware.

3
00:00:09,540 --> 00:00:11,360
So as we can see in this picture.

4
00:00:11,670 --> 00:00:19,320
This is a model of oversimple narrowing where we have inputs from X1, x2 to X and and each input has

5
00:00:19,320 --> 00:00:25,350
a specific rate, not only one W two W and are showing our weight.

6
00:00:25,830 --> 00:00:34,980
All of them are going to air some signal with B, which is forever bias and the bias will say our system.

7
00:00:34,980 --> 00:00:40,900
How is Twiggy's it after that of our next signal is going to activation function.

8
00:00:40,920 --> 00:00:45,050
This is where the magic will happen and finally we can have over output.

9
00:00:45,570 --> 00:00:48,690
So let's take a look at this model here.

10
00:00:48,690 --> 00:00:50,070
We have over inputs.

11
00:00:50,070 --> 00:00:56,400
Let's call them X1 here X two and then two X in.

12
00:00:58,260 --> 00:01:00,350
Each of them has a specific weight.

13
00:01:00,720 --> 00:01:02,970
We are showing the weight with a resistor.

14
00:01:03,000 --> 00:01:04,560
These are our resistors.

15
00:01:05,130 --> 00:01:11,520
W one, W two or R one are two, two RF and which are of our resistors.

16
00:01:12,030 --> 00:01:16,260
And finally we will just feed them to an up EMP.

17
00:01:17,850 --> 00:01:25,940
Here we have over op amp this up AMP will act as our activation function.

18
00:01:25,950 --> 00:01:38,370
So this is our F function and then F of net of course, and then here as a bias, just look at this

19
00:01:38,370 --> 00:01:39,660
model as a bias.

20
00:01:39,660 --> 00:01:45,870
We will just connect it to the battery and then we can just adjust the voltage of the battery to have

21
00:01:45,870 --> 00:01:48,210
a bias that we are looking for.

22
00:01:48,450 --> 00:01:51,630
And finally, this system will give us an output of AC.

23
00:01:52,110 --> 00:01:58,080
You can simply design your neural network using different softwares.

24
00:01:58,080 --> 00:02:03,450
And then if you want to implemented using hardware, you can just use this model.

25
00:02:04,650 --> 00:02:10,340
There are several advantages of using an analogue system compared to the digital system.

26
00:02:10,350 --> 00:02:17,700
One of them would be unadopt systems are much more faster than digital systems, so they can respond

27
00:02:17,790 --> 00:02:19,170
very faster.

28
00:02:19,170 --> 00:02:22,530
But of course there are some limitation for our knowledge of systems.

29
00:02:22,860 --> 00:02:29,100
They are not extensible and we can just use them for a specific purposes.