NDT Classifiers

Max Correlation Coefficient

Poisson Naive Bayes

For Now, Please refer to the script used in the Max Correlation Coefficient

LibSVM

Draft

SVM Notes

-Use RBF

1)Transform data to format of SVM Package (This has already been done as well) 2)Conduct Simple Scaling of data (toolbox already does this! Use the RBF kernel Use CV to find best C and gamma parameters Use these parameters to train the WHOLE training set Test it

-Linear kernel is special case of RBF -RBF has fewer numerical difficulties -I have a small number of features thus RBF kernel is suitable

-A ‘grid search’ on C and gamma using CV (HOW???) -Various pairs are tried and the one with the best CV accuracy is picked. -Use exponentially growing sequences of C and gamma (2-5,2-3,…2^15, and for gamma = 2^-15,-13…,2^3)