The following content has been provided by the University of Erlangen-Nürnberg.

Hello everybody. So as the semester is slowly winding down and the last weeks are approaching,

also our lecture comes to an end, we're going to talk about some fundamentals of automatic decision making today,

we hinted at last week and this is under the headline fundamentals of pattern recognition and machine learning,

which are kind of two words for similar things.

Before we step into that, I want to recap as very often a little bit what we talked about last week.

So please for a couple of minutes take out your smartphones, we'll tweet back again today,

and go to www.tweetback.de and use the session code x79 for answering the following few tweet back questions.

So session code is x79.

And the first question is while you're opening, how would you fill the gaps?

A waveform is the da-da of a signal, it shows the changes in over a certain amount of whatever,

and the ECG signal has simple and recognizable, which are modified by pathological processes.

And people are still looking at their smartphone screens, so let's wait a few more seconds.

Okay, so first answer could either be wavelet graph frequency or graph, so wavelet graph of frequency.

So it's, wavelet would be a little bit too small, frequency doesn't even fit, and so graph remains,

so it's either answer B or D.

Now let's see first what you guys said, so one person mistyped and all the others are correct.

So the correct answer is answer B.

And I'll start by another quiz already, but just to give a few more words on that.

So it shows the changes in amplitude or time, so obviously this is amplitude over a certain amount of time,

and therefore answer B is correct.

Very simple, so warm up.

Okay, so the question I already opened, second question, which of the following measures from a heartbeat

are useful for characterizing heartbeat waveforms?

Could be QRS with T amplitude, RR interval, PQ interval, RS amplitude.

Is it none? Is it the given selection in answer B or C? Or is it answer D?

Hi, good afternoon. We are doing a Tweet back, so if you want you can join by using a smartphone

for answering the questions here.

Oh, I didn't start it, I'm sorry for that. So let's start it, and now you should be able to vote.

I have, as you might know, I have a lecture on usability, so we just detected a usability issue.

If you want to learn more, you can come back in summer semester,

with HCI, Human-Computer Interaction Lecture again.

Okay, I filled the break, so let's see what you voted, and you're still voting.

So a couple of people say B is correct, C is correct, D is correct, nobody votes for A.

Oh, this is dynamically changing.

So last time we had 15 votes, now we have also 15 votes, 16, 17.

So you guys have to decide, I mean, who's going to be a millionaire, this wouldn't help me.

Actually, I would say all of them are helpful.

So the question is so unspecifically asked, just what is useful for characterizing heartbeat waveforms,

and you can generally use everything.

I wouldn't say that any of them is not helpful for characterizing waveforms in general.

Now if you look at specific things, like for example, you want to quantify,

or you want to tell a medical expert something about the changes in,

or the risk factors for myocardial infarction, you might want to look at the isoelectricity of,

for example, the ST segment, and then, okay, cool, we don't have the ST segment here,

so I picked the right thing, but it might also be the PQ interval that you might want to look at,

so the isoelectricity of that, and then maybe it's not so interesting to, for example,

look at the RR interval, just an example.

But we will learn today that in machine learning, if you implemented your process correctly,

then it doesn't matter, because you typically have a feature selection process,