●</●

●</●

●</●

"...Welcome everybody to the Monday session.

Mondays, we usually have 90 minutes, so I take the first

five to ten minutes to build up the big picture

to show you the Mindmap,

such that you can follow up the party line

or the story line of the Lecture.

And we are here talking about, what is this R here?

That's not good, pattern analysis.

And pattern analysis is basically

about one core concept that we have seen

in winter semesters already,

and that's the Bayesian decision rule

BASIAN-CLASSIFIER.

The BASIAN-CLASSIFIER has a very nice property.

The BASIAN-CLASSIFIER applies a decision rule

that turns out to be optimal with respect

to the zero one loss or cost function.

The BASIAN- deciddient rule is

we decide for the class that maximizes

what, that maximizes what Rotard.

Families.

Yes, but how do we decide.

need for af goal.

Yes.

We maximize the.

a posteriori likelihood.

P Fr Y

given feature vector X

and we maximize over why we have seen.

In this lecture last week.

Again, this is an optimal decision rule

according to the average loss.

If you apply this decision rule,

you will end up with the lowest

possible average cost of your classifier.

If you change the cost function,

things look very much different.

But basically, if you say

correct decision is for free,

if you do something the wrong way,

then you have to pay one dollar then

is the currency in your home country?

Rupees.

Rupees, or you have to pay one rupee for a wrong decision.

Yeah.

If you have this type of cost function,

then independent of the place in the world you are,