1
00:00:00,600 --> 00:00:07,670
This is a block diagram off of a multilayer Perceptron where we have a picture of X for our input,

2
00:00:08,010 --> 00:00:16,620
we have here twin neurons and one into an entry, the output of this neuron is Y one, the output of

3
00:00:16,620 --> 00:00:18,240
neuron two is Y two.

4
00:00:18,240 --> 00:00:22,650
And similarly, we have Z for our output of these.

5
00:00:23,610 --> 00:00:24,210
System.

6
00:00:25,680 --> 00:00:35,660
Let me write down why one as the function of Ivanoff here we have w transpose w one transpose, we're

7
00:00:35,670 --> 00:00:42,740
going to just label it with W one X plus V A bias of B for this now.

8
00:00:42,750 --> 00:00:46,770
And this is for the first Nahrin and one to show the output of first Nahrin.

9
00:00:47,130 --> 00:00:55,350
For the second number we have Y two as an output two and this one is W to transpose X.

10
00:00:57,140 --> 00:01:03,620
Plus B, let's just label them B one and B to.

11
00:01:05,830 --> 00:01:16,660
Now, for the Z function, we can also write Z is equal to F of three this time W three, transpose

12
00:01:17,260 --> 00:01:25,440
X or let's say why it is better to write it with Y plus MI three and four F of three.

13
00:01:25,540 --> 00:01:42,640
We can write W three one y one plus W three to buy one y two this time plus B three.

14
00:01:43,440 --> 00:01:52,180
So we just expand this W three Y and now here we can continue and write as Z F of three.

15
00:01:52,180 --> 00:02:05,380
If we just expanded here I can write it as W three one f one F funny's W one transpose X plus B one

16
00:02:05,830 --> 00:02:17,080
and then it just replace it with its value plus W twe two, F of two and F two.

17
00:02:17,080 --> 00:02:28,650
Here is W to transpose X plus B to just don't forget this Vetri plus B tree and just close to book.

18
00:02:29,830 --> 00:02:34,500
So we see that multi-layer Perceptron is nothing but a function.

19
00:02:35,740 --> 00:02:41,130
But why did we do that and why do we need to take a look at this one in this way?

20
00:02:41,680 --> 00:02:47,230
If you're using a software, if we are using different techniques of software, we don't really need

21
00:02:47,230 --> 00:02:48,820
to write down this function.

22
00:02:49,090 --> 00:02:55,630
But it's a very good example to understand what is a simple Neren and what is this algorithm that later

23
00:02:55,630 --> 00:02:59,740
we are going to work with how they work and what our day basson.

24
00:03:00,190 --> 00:03:07,510
If you are really looking into finding a mathematical function, we have different tools as simple tools

25
00:03:07,510 --> 00:03:14,890
like a curving fitting tools where you can just walk me to functions and right on separating two sets

26
00:03:14,890 --> 00:03:15,490
of data.

27
00:03:16,480 --> 00:03:22,390
I hope that you can understand now the fundamental of Narron and what is a neuron?

28
00:03:22,390 --> 00:03:28,030
How can we represent a narrow made a mathematical function and hard working?

29
00:03:28,420 --> 00:03:32,740
And why do we need to use multilayer Perceptron in the next session?

30
00:03:32,740 --> 00:03:37,000
We're going to discuss about hardware implementation often never.