1
00:00:00,750 --> 00:00:09,210
But let's have a model for a simple artificial Nahrin here resolver example about some patients which

2
00:00:09,210 --> 00:00:12,080
we want to see if they are diabetic's or not.

3
00:00:12,450 --> 00:00:21,450
In this example, we just a function for classification exposed behind plus equal to zero.

4
00:00:21,930 --> 00:00:28,480
Our inputs here are x1 x2 x3 through X often.

5
00:00:29,820 --> 00:00:37,610
In reality, there are more than one parameter to help us making decision for a classification.

6
00:00:38,040 --> 00:00:40,890
So it's not only one factor here.

7
00:00:40,900 --> 00:00:48,690
There are actually more elements that have some effect on our decision for drawing this classifier of

8
00:00:48,690 --> 00:00:49,110
line.

9
00:00:50,440 --> 00:00:53,640
But you might also ask, so what happened to the B here?

10
00:00:53,640 --> 00:00:55,830
We have B of Y.

11
00:00:56,160 --> 00:01:00,330
B is an element which doesn't have any affect in our decision.

12
00:01:01,110 --> 00:01:04,970
Let's say me, is how tough the patient.

13
00:01:04,980 --> 00:01:11,430
It doesn't have really affect to our making decision for if the person is diabetic or not.

14
00:01:12,000 --> 00:01:16,560
So some data, they don't have any effect on our output.

15
00:01:16,810 --> 00:01:21,510
We have this Y here, but the weight of Y is just zero.

16
00:01:21,800 --> 00:01:26,220
We don't consider it is just like V is zero.

17
00:01:26,460 --> 00:01:33,940
If means equal to zero, then it means the effect of D Y in our output would be zero.

18
00:01:34,440 --> 00:01:35,610
Take a look at this one here.

19
00:01:35,610 --> 00:01:38,520
We have different weight for each input.

20
00:01:38,850 --> 00:01:46,890
It means each input has a specific effect on the output and we show it with W one W two true.

21
00:01:46,890 --> 00:01:56,300
W if I send all of them to a sigma and some sum of all of them would give us an output function.

22
00:01:57,510 --> 00:02:03,420
But before that, here we have some other factor, a very important factor which we call lead by us

23
00:02:03,420 --> 00:02:04,520
off of our system.

24
00:02:05,190 --> 00:02:08,040
We can't show it with B or any other LATRA.

25
00:02:08,040 --> 00:02:11,040
It's usually actually in most of the textbook.

26
00:02:11,040 --> 00:02:12,090
It's just with B.

27
00:02:12,540 --> 00:02:15,120
B is a standard for a bias of our system.

28
00:02:15,140 --> 00:02:17,280
But why do we call it bias?

29
00:02:18,000 --> 00:02:24,720
You have X, me Y plus C, what is the effect of C in our system?

30
00:02:24,860 --> 00:02:36,270
If I set C to the zero then I will have a line off A X plus B Y equals two zero X plus B Y equals to

31
00:02:36,270 --> 00:02:40,200
zero is just going from this line.

32
00:02:40,290 --> 00:02:42,390
It should be somewhere around here.

33
00:02:42,720 --> 00:02:44,580
So this is a straight line.

34
00:02:45,480 --> 00:02:47,450
Here is our would be our line.

35
00:02:47,460 --> 00:02:54,190
It means we're so strict, the doctor is so historic and this patient who was healthy before that,

36
00:02:54,420 --> 00:02:57,290
we are now considering as a diabetic as well.

37
00:02:57,840 --> 00:03:01,140
So D. C. is showing the bias of the system.

38
00:03:01,140 --> 00:03:02,730
And how is Treweek is a doctor.

39
00:03:03,000 --> 00:03:08,700
And if we just increase the amount of C, then we will have to line somewhere here.

40
00:03:08,700 --> 00:03:12,240
It means all the patients now are healthy.

41
00:03:12,810 --> 00:03:16,220
This bias will show how streak is over a system.

42
00:03:16,980 --> 00:03:24,360
Of course, it has also some effect and we can show this effect by W zero and then some of all of them

43
00:03:24,360 --> 00:03:30,320
is giving us an output function, which we call it usually as a activation function.

44
00:03:30,690 --> 00:03:40,260
This is our net signal and that signal is giving us cell activation function that we just erase it all

45
00:03:41,880 --> 00:03:42,150
here.

46
00:03:42,150 --> 00:03:48,480
I can say they would be equal to F off net.

47
00:03:50,170 --> 00:03:56,800
This next signal is equal over by us B plus.

48
00:03:58,550 --> 00:04:14,060
Some way, W one, x one, and here we have W two, x two plus just goes to W of an X of N.

49
00:04:16,140 --> 00:04:26,850
Or in other hand, we can say that the next signal is just equal to be plus the sigma of W i.

50
00:04:28,410 --> 00:04:29,270
X of my.

51
00:04:31,000 --> 00:04:41,840
Here is a model, a very simple artificial narrowing, which we know as my clock, Pete Neron, my clock,

52
00:04:41,840 --> 00:04:48,230
Pete was the first model of artificial Neron and it's a very simple Neren to show our outputs.

53
00:04:48,260 --> 00:04:56,030
Let's take a look at these actual Neron in this and we can say, oh, we can see also we have some inputs.

54
00:04:56,270 --> 00:04:58,400
Here are our axes, the inputs.

55
00:04:58,730 --> 00:05:02,340
Each of them has a specific effect on the output.

56
00:05:02,570 --> 00:05:04,010
We show this effect.

57
00:05:04,010 --> 00:05:04,850
How important?

58
00:05:04,850 --> 00:05:07,680
Because some details are more important than other data.

59
00:05:07,700 --> 00:05:10,310
So how we you show this importance.

60
00:05:10,310 --> 00:05:12,110
We show it with W here.

61
00:05:12,110 --> 00:05:14,360
We have different inputs.

62
00:05:14,360 --> 00:05:16,310
Each of them has a specific rate.

63
00:05:16,580 --> 00:05:19,670
And then here we do some process.

64
00:05:20,240 --> 00:05:29,330
And finally after this process, we send a net signal to the output and here are our Z function and

65
00:05:29,330 --> 00:05:30,650
our output.