Okay, welcome everyone to the next lecture of Introduction to Machine Learning. Let's start

again with a small quiz. I hope you can see the link here. I already started it so you should be

able to directly vote. So the first question is about ordering. So you have

given some images. So these are the AMS and you should basically tell which of

the spectra belongs to which letter. So order them in the correct order. So is

this one here which one belongs to this which letter and so on. Two people

already voted. For those who come a little bit late can still vote.

Okay and the code is zq84. Okay let's see what you got. Oh only two? You are more

than two. Come on. So what do you think?

So you remember that you can see here these lines and this plot here in these

plots. Okay seems you're not so motivated today or only two people are

motivated. Doesn't it not work? So again the code. Well okay then let's continue.

See what you got. So I know okay we have many more. 15. That doesn't say two. Okay great.

So a little bit more of the half have you tried. So let's go through it.

So this one's here. So the first one it has very round shape in different

circles. So in different Rade. So because of this roundedness it belongs to

this one here. So there are no clear I mean this you see that there is a more

straight one and a slight angle here. This is probably because of these and

that you have some frequencies here. But overall it's definitely in contrast to

the other two much more rounded shapes. So S1 would belong here to the S. And now

we have the middle one. So that basically you can see directly which

one is which because this one here it has a straight vertical line and so it

must belong to this A because it has this one in contrast to the M. And this

one here has quite vertical lines here so it has horizontal lines in the

spectrum. So this S2 would be then M. And of course the other two bright lines

are related to these angles here and here to these lines. So they are all

always orthogonal in the spectrum. Okay let's go to the next question. So

here some multiple choices are again possible and the question is yeah you

have to select the correct answer. So it's a little bit about properties of

different orthogonal bases.

Okay 12 of the 15 already submitted an answer. I think we can check what you got.

So Fourier coefficients. So the difference here between A and B is just that the

Fourier coefficients are translation invariant or the magnitude values of the

Fourier coefficients are translation invariant. So here A per se would not be

translation invariant. So only if you really take the magnitude of the

real and imaginary value this one would then or these values would then

be translation invariant. So then not many people voted for that. This is good

because this is not true. They span an orthogonal basis but not an

orthonormal basis. So they have a factor there the features. In contrast to the

Walsh features, the Walsh features they span really an orthonormal basis. And the

last one here the discrete Walsh functions if they can be recursively or

defined by the Hadamard transformation and this is of course true and this is

what we learned that we have this Hadamard, small Hadamard matrix and with

the Kronecker product you can basically build increasing numbers of Walsh

functions. So it's a quite nice way of creating these discrete Walsh functions.

Okay this was it already for today. A small mini quiz but before we switch

over to the lecture do you have any any questions regarding the quiz or

regarding the lecture, regarding the exercise? Is there something we should

discuss here or is there something you would like to change, improve? No, not here

right now. Okay then let's start with audio and speech features. I think