Options

# JmySVMLearner C parameter question

niccolo_camarli
Member Posts:

**3**Contributor I
Dear all,

I'm trying to use JmySVMLearner, but i have question that i'm not able to answer by myself.

One of the parameters of this operator is C, what happens if i select a C value= 0.0 ?does the algorithm switch to the linear separable case of the SVM (no constraint for the Langrange Multipliers) ?

Tank You in advance for your help

I'm trying to use JmySVMLearner, but i have question that i'm not able to answer by myself.

One of the parameters of this operator is C, what happens if i select a C value= 0.0 ?does the algorithm switch to the linear separable case of the SVM (no constraint for the Langrange Multipliers) ?

Tank You in advance for your help

0

## Answers

439Mavensee the mySVM documentation:

C: the SVM complexity constant. If not set, 1/avg(K(x,x)) is used.

Here, 0 means "not set" :-)

If you set the log level in the process root operator low enough, the SVM will log the value that is actually used.

Cheers,

Simon

3Contributor IBut where this 1/avg(K(x,x)) comes from?

I found this value also on the website of Stefan Rüping (Mysvm ) and also on SVMlight website, could you give me some reference that explains the reason of this C guess?

Tank You

439Mavenhave a look at the references below. Thinking about it, it occurs to me that the formula I copied from the mySVM documentation may be wrong. Please have a look at the literature to be sure.

Cheers,

Simon

http://www-stat.stanford.edu/~tibs/ElemStatLearn/

Hastie, T., Rosset, S., Tibshirani, R. and Zhu, J. (2004). The entire regularization

path for the support vector machine, Journal of Machine

Learning Research 5: 1391-1415.