You are on page 1of 6

SVM: Background Study

1
Background Study
Support Vector Machine (SVM):
• Follows structural risk minimization (SRM) principle.
• Finds an optimum classifying hyperplane: x   b  0
t

so that the classifier becomes: sign ( x t


  b)
Primal Optimization Problem of SVM:
1
min ||  ||2
2
s.t. yi ( xit   b)  1, i  1,2, , l

Equivalent Dual QPP of SVM:


1 l l l
   i yi xi x j y j j    i
t
min
2 1 j 1
i i 1
l
s.t.  yi i  0
i 1

0  i 2
Background Study
Support Vector Machine (SVM):
Support Vectors are the training data samples such that their corresponding
i  0
Example:

3
Background Study
Support Vector Machine (SVM):
Support Vectors are the training data samples such that their corresponding
i  0
Example:

4
Background Study
Support Vector Machine (SVM):

 Optimization Problem of SVM with error tolerance:


l
1
min
2
||  ||2 C 
i 1
i

s.t. yi ( xit   b)  1   i ,
i  0
 Equivalent dual QPP:

l l l
1
min
2 i 1 j 1
 i yi xit x j y j j  
i 1
i

s.t. y
i 1
i i 0

0  i  C 5
Thank You!

You might also like