2013B+2013C

2013B+2013C

Dor (guest) 27 Jan 2017 14:44

Hi,
I got 3 main questions:
1) in 2013B I didnt get the answer to 5a. Why is the answer yi/xi-a and not y-xa?

Becuase of that, I didnt know how to answer 5a in 2013C.

2)in 2013C I didnt get the answer to 7.
How the regression line is computed? and how the PCA does?
Did we learn it?

Becuase of that, I didnt know how to answer 7 in 2013C.

3)What is the answer to:

Thanks.

Re: 2013B+2013C

Regev Schweiger 30 Jan 2017 15:19

1. I can't debug your proof from afar. Write the likelihood of the data explicitly (based on the pdf of the normal distribution). Remove constants that do not depend on a - this is what you get.

2. You can calculate them simply apply the algorithms we learned in class (pick some parameters for the points). In that year, there was slightly more focus on that subject that made this question easier, but it's completely solvable.

3. If I recall, this is answered in some other exam's solution - am I wrong?

Dor (guest) 30 Jan 2017 19:57

Desictoin tree yes.
Soft SVM no.
If we give errors and than transform the data, it can cuase more errors.
So What will be the answers to ax,Dx,Ux?

Regev Schweiger 31 Jan 2017 08:49

A general Dx might change it, ax and Ux will not.

Dor (guest) 31 Jan 2017 10:59

Thanks.
How about k means?

Regev Schweiger 31 Jan 2017 11:13

Same thing.

Dor (guest) 31 Jan 2017 11:31

mutiply by D-Why does shrinking the axis can change the output?
The closest points will be still the closest

student (guest) 31 Jan 2017 11:35

יכול להיות שנקודות היו קרובות אחת לשניה בציר מסוים,
אבל אם הציר הזה מוכפל פי 1000 וציר אחר לא, הנקודות הקרובות יכולות להשתנות

Roi (guest) 31 Jan 2017 19:11

I disagree about ax. I think it changes the dinamic of "how soft is the margin" - similar to changing c. (because the Xi_i don't scale with the sampling points.)
If i take a very large a , the resulting classifier will be almost hard-margin.
If i take a very small a, the resulting classifier will care very little about mistakes

Am I wrong in this?

Regev Schweiger 31 Jan 2017 19:30

Actually, you are correct, if the question refers to the same C. Thanks.

Dor (guest) 28 Jan 2017 08:26

Another question:
Why in 2013B question 6 the margin is 2sqrt(sum of alphas)?

Regev Schweiger 30 Jan 2017 15:20

Based on an equality we did not learn this year.