matlabmachine-learningsvmlibsvm

Data format for Libsvm SVR training in Matlab

I have two questions related to data input in LIBSVM.

  1. Do I need to format the data into sparse format to input svr libsvm in matlab?
  2. Do I need to normalize the data before I input them to the trainer?

I am training svr without doing any of these and I am getting the same results even when formatted. As mentioned in the documentation of libsvm the data formatting is used when we do it for OCTAVE, since every thing is ran automatically by just running train.py and test.py. But in matlab I am not sure.

Can someone clarify?

Solution

  • 1) This may be a good idea to do, to be more effective.

    2) It is recommended but not required.

    But like I said in comment here, LibSVM matlab use to call C functions. So you can refer to main documentation of libSVM for those kind of question.