Prüfung 26.01.2012

[Seimuun]

questions from memory:

1) data preprocessing - what is it, why?
2) bagging - boosting? what is it, which algorithms use it?
3) what are the concepts "linear separation" & "soft margins"? which classifiers use it?
4) what is the concept "landmarking"
5) practical: decision tree & bayes net (design & calculate bayes net, than classify a new instance, what happens if node "debt" is instantiated)

since as far as i can remember we never actually built a bayes net from scratch (with a dataset and probabilities), the 5th question was really hard for me...

[emptyvi]

As question 5) (with k-NN ensemble instead of decision tree) came on this test (30.04.2012) too, but even after the test, I have no idea how to solve something like this.. could anyone explain it to me?

In our example, a table with features name{some names}, eye colour{brown, blue, green}, height{short, tall}, handsome{yes, no}, sex{male, female}, Likes_soccer{yes, no} and class date{yes, no} was given. (I dropped feature "name" as I don't think, it would be appropriate as it is a unique identifier for each dataset.)

regards,
emptyvi