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

I have a couple of announcements regarding the oral exam.

So there will be an oral exam. We have an exception.

We have for this semester an oral exam, although we have so many applications.

We are not sure if we will have that in the future.

But this year it will be an oral exam for sure.

And you can register for the oral examination in our secretary's office.

There are different appointments available.

So depending on the date that you pick for your oral examination,

you may either end up with Stefan Stadel as the examiner or with Professor Hornegger as the examiner.

So you can decide that when you apply for the oral examination.

You can go there to our secretary's office starting from tomorrow morning.

This afternoon nobody is in, so you can't register right now.

Are there any more questions about examination dates?

So the dates, I only know the second term, so the second term will be from the 21st of March until the 12th of April.

I know this is the date for the...

No, I don't know the specific dates.

And we had to share the dates depending on who is available and who isn't.

Because we also have to take care how many people can actually be examined at the same time.

Because we don't have that many rooms for oral examinations.

So you have to go there and ask.

There are enough appointments for everybody to get examined.

Hey, welcome.

Okay, so let's get back to pattern recognition.

So last time you heard something about the EM algorithm and how to solve problems using the EM algorithm, is that right?

So what did you hear?

I haven't been there, I haven't watched the lecture.

So maybe somebody can say a few words, what you learned.

Who has been here?

Who has been here at all?

Nobody. Nobody remembers what the EM algorithm is.

Okay, there's still people coming.

Do you remember what the EM algorithm is?

No? Have you been here last week?

No?

That's bold.

Good for you.

Okay, what's EM? What does EM stand for?

Who remembers that? What does EM stand for?

Yes?

Expectation-maximization.

Yes, exactly. So there's...

Yeah, so what is the thing? When do we like to apply the EM algorithm? In what cases?

There was something with missing information.

Yeah, so if there is something hidden, some hidden variables, something we cannot observe.

Ah, okay. We will summarize it tomorrow again.

But we remembered it's expectation-maximization.

For the EM algorithm, you have looked at that you can actually prove convergence.

So if you do a step, an EM step, your log likelihood function improves, which gives you a better result if you keep on repeating the EM algorithm.

And doing so, you can estimate parameters with hidden variables.