Support Vector Machines Flashcards
(9 cards)
1
Q
What is the margin?
A
2/norm(w)
2
Q
How do we maximize the margin?
A
Minimize norm(w)
3
Q
What is a hyperplane?
A
Regions, made by a seperation plane
4
Q
Which data points define a seperating hyperplane?
A
The nearest vectors for which and only these:
y*(w^Tx+b)=1
These point lie directly on the margin
5
Q
What are support vectors?
A
Points on the margin
6
Q
Is every set of data linearly seperable?
A
No, because of that we allow the SVM to violate some constraints
7
Q
What is the slack variable?
A
Describes the extent to which the respective contraint is violated
8
Q
Soft Margin and hard margin
A
With Hard margin, we do not allow constrainst and with soft margin we allow constraints.
In hard margin -> no slack variable
In soft margin -> with slack variable
9
Q
A
