What are they best optimized parameters to be used in svm for binary classification and linear kernel e.g values of -c, -b, -h, -s etc to get a best separating clusters of svm with separating hyperplane.
please do some research first. There are so many ressources out there with good explanations, e.g.
http://www.csie.ntu.edu.tw/~cjlin/libsvm/
http://svmlight.joachims.org/
Dear Sheema.
In SVM the problem is finding the right kernel, after finding it another problem is finding the right set of hyperparameters like a,c,h in linear kernel. For finding the parameters there are well defined methods like:
1.Grid Search
2.Evolutionary methods like PSO
3.Other Optimization methods.
Parameter C is called soft margin constant and is very important:
1.Big C indicates we want to have a high train accuracy and consequently lower test accuracy.
2.Small C shows that we want to have a tolerated amount of error in training for the sake of lower error for test.
For more info please see link below:
http://scikit-learn.org/stable/modules/svm.html
This is the shrinking parameter. For most applications you want to use the default value. More infos in chapter 6:
http://dsec.pku.edu.cn/~jinlong/pksvm/An_Summary_of_svm.pdf
- Badges
- Science topic