HTTPS.

Test test.

Okay, sorry for the little delay today.

I had to record the lecture by myself again

because the official recording is not here today.

I hope everything works.

I just set it out.

But yeah, let's hope for best.

First things first, I already transferred the points

regarding the submission two to student.

If you participated in that lecture, in that submission,

please check whether your points are correctly transferred

so we can correct it if they are wrong.

But I've seen already more than 30 people

have passed the 72 points requirement for the mock exam.

So that's pretty good, I guess,

before the last submission starts today even.

Okay, with regard to the lecture,

we stopped at the last subsection of clustering last time.

Have there been any questions regarding clustering

at all, all the topics before

the evaluation of clustering section?

Have there been any questions regarding that?

Okay.

Then let's go to the last subsection of clustering,

evaluation of clustering.

Of course, we have to check

whether our clustering results are good

and we should be able to check it

even if we can't visualize the clusters well.

The most clusters we talked about in the exercises,

we have always been able to just

visually evaluate our clusters.

For example, if we have six points

and we have two clusters, that seems to be fine, of course.

However, if we have not coordinates, for example,

but we have distances between names, let's go with names,

it's harder to assess whether our clusters are very good.

And if you remember the last lecture

or even the second to last lecture,

if I remember correctly,

then we talked about what are good clusters.

We want a high intra-class similarity

and a low inter-class similarity.

So high similarity within the same cluster

while we have high difference, high distance

between points of different clusters.

And basically, that's what we want to check,

we want to assess.

For this, we have different methods, again.