Hello everybody, my name is Stefan and I will be giving together with Dahlia this year's

pattern recognition exercises.

In this video I will present you briefly what you can do with this pattern recognition toolbox

that we already showed you.

As you might have already read and understood, you can download the pattern recognition toolbox

and install all the dependencies with this string here.

So depending on which Linux or Windows you are using, you might need to use pip2 or pip3.

For me it doesn't matter here in this situation.

And when you are starting the pattern recognition toolbox, you can do this by running python

toolbox and then py classification toolbox.

You will still see some serrations that will be looking like this.

And basically what we are going to be doing this semester is looking at features in multi-dimensional

spaces.

And what the py classification toolbox essentially does is it visualizes a two-dimensional space,

the only space that we can visualize on the screen.

And we will use it to implement different classification algorithms and visualize them.

So what you can see right here when you are going to be starting it, you will see only

a pair of dots.

You can delete them by just saying here new and you can also create new points by clicking

on this button here.

When you are doing this, you can select a class for each of the dots and set them with

the mouse.

And depending on which class number you are using, the dots will get a different color.

For example green for class 3 and red for class 1 in this case.

I used here class 6 for blue and I just used this class numbers because they had colors

that are easily to distinguish here on the screen.

When you are going back to this button, then you can again move, zoom, pan and so on.

And yes, this is essentially a functionality.

You can also use this button here to create a Gaussian distribution also with selecting

the class.

But in the beginning we probably don't need this functionality because as you can see

it creates a lot of samples and since our algorithm will be very slow, it can take a

lot of time to calculate with that high amount of points.

So as I said, you can just place dots here and also delete them again.

With this button here.

And basically these are the samples that you can...

So this was the first step to just set samples and just create some random problem situations

for your algorithm.

Our first problem will be to implement a K-Nearest neighbor classifier.

And I already implemented it so I can go to classification, K and N for K-Nearest neighbor.

Then I click K and N and then I can just select a number of neighbors.

Like we will get what this actually means in a second.

And then click classify and what it will be doing is it will take each and every screen

point and will assign a class and class will be just represented by a color again.

And when you implemented your algorithm color-rectly, you should see like a decision boundary with

your 2D space divided into different classification areas and you should be able to see clearly

decision boundary.

So yeah, I was talking about the K-Nearest neighbor classifier.

What does this actually mean?

So when we are going back to the introduction of pattern recognition, what you saw in the