﻿1
00:00:00,533 --> 00:00:02,633
Hello and welcome back to the course
on Machine Learning.

2
00:00:02,633 --> 00:00:04,700
I hope you enjoyed the previous tutorials.

3
00:00:04,700 --> 00:00:09,100
And today we're going to talk about
the different types of kernel functions.

4
00:00:09,433 --> 00:00:14,066
So the final thing that you need to know
about the kernel SVM is that the radial

5
00:00:14,100 --> 00:00:18,666
basis function, which is also called
the Gaussian function,

6
00:00:18,900 --> 00:00:23,200
is a not the only kernel function
that is used in this method.

7
00:00:23,900 --> 00:00:25,000
so let's have a look at a couple.

8
00:00:25,000 --> 00:00:29,266
So here we've got the Gaussian or the RBF
kernel, which we've already spoken about.

9
00:00:29,600 --> 00:00:33,600
Then there's also a popular choice
which is a sigmoid kernel.

10
00:00:34,266 --> 00:00:35,633
And that's the formula on the right.

11
00:00:35,633 --> 00:00:40,100
It's in a bit of a different notation,
but the essence is the same is that you

12
00:00:40,100 --> 00:00:44,033
still select a, landmark and then,

13
00:00:44,300 --> 00:00:47,733
then from there,
depending on the distance,

14
00:00:47,766 --> 00:00:52,100
the landmark, different results
will occur.

15
00:00:52,100 --> 00:00:56,600
But in this case, as you can see,
this kernel function is directional.

16
00:00:56,600 --> 00:01:00,933
So anything, even like just looking at it
in this, two

17
00:01:00,933 --> 00:01:02,866
dimensional space is just a projection.

18
00:01:02,866 --> 00:01:07,833
You can see that anything to the right
will, be automatically right away.

19
00:01:07,833 --> 00:01:09,366
will have a high value.

20
00:01:09,366 --> 00:01:11,466
So it will be included
in your classification.

21
00:01:11,466 --> 00:01:12,866
Anything to the left will be excluded.

22
00:01:12,866 --> 00:01:15,466
So sometimes
you might require these situations.

23
00:01:15,466 --> 00:01:18,733
So maybe if you look at a two dimensional
space, just imagine those points

24
00:01:18,733 --> 00:01:23,300
that we had points on on
one side are not in your classification.

25
00:01:23,300 --> 00:01:25,733
On the other side,
are you in your classification.

26
00:01:25,733 --> 00:01:29,033
You want to somehow, outline

27
00:01:29,033 --> 00:01:32,866
that decision boundary, or, highlight

28
00:01:32,866 --> 00:01:37,333
that as, points past a certain points are
should be in your classification.

29
00:01:37,333 --> 00:01:40,433
Then a sigmoid kernel is a popular choice.

30
00:01:41,333 --> 00:01:45,433
Also, we've got, polynomial kernels,
which are also popular.

31
00:01:46,266 --> 00:01:48,000
one of the more popular choices.

32
00:01:48,000 --> 00:01:51,800
And, here
you can have, like, a, polling them,

33
00:01:52,100 --> 00:01:55,433
which dictates how your kernel behaves.

34
00:01:55,433 --> 00:01:58,066
And, again,
you can go into more detail on that.

35
00:01:58,066 --> 00:02:01,800
So this tutorial is not for us to discuss

36
00:02:01,933 --> 00:02:05,066
the specifics
of each one of these, kernels, but

37
00:02:05,066 --> 00:02:11,800
just to, show you that they exist and,
help you be aware of the different types.

38
00:02:11,800 --> 00:02:15,966
And so these are the most popular choices
the Gaussian, RBF,

39
00:02:15,966 --> 00:02:17,200
sigmoid and polynomial.

40
00:02:17,200 --> 00:02:18,700
Hope you enjoyed,

41
00:02:18,700 --> 00:02:22,166
learning about the kernel SVM
and I look forward to see you next time.

42
00:02:22,233 --> 00:02:23,966
Until then, happy analyzing.
