※ 引述《rafan (@_@"""")》之铭言：
: I think Dr. Lin said that you should use the same factor
: for scaling training and testing data, i.e.,
: svm-scale -s range ijcnn1 > a
: svm-scale -r range ijcnn1.t > b
: Is that true?
yes, and scale it to [0,1]

thanks