changes.txt

Support Vector Machines is a powerful methodology for solving problems in nonlinear classification a

MATLAB Interface for LIBSVM
===========================


Version 1.2 (12-Sep-2005)
-------------------------

- Now using LIBSVM 2.8.
- Alternative output formats (as vectors).
- Added support for arbitrary class labels (not only +1/-1).
- Added support for probability estimates in libsvmsim.
- New option 'prob' to compute probability information.
- Cache size can be given in kB, MB, or rows.
- verbose() split into dumpOptions() and dumpSolution().
- dumpSolution() computes epsilon for NU-SVR.
- Small improvements all over the code.


Version 1.1 (28-Feb-2005)
-------------------------

- Now using LIBSVM 2.71.
- Fixed bug which produced wrong sign of the decision function.
- Kernel struct and function getKernel() rewritten.
- Added tanh kernel and offset term for polynomial kernel.
- Added data weighting for classification.
- Options can now be passed also as key/value pairs.
- Added support for nu-SVMs and one-class SVMs.
- Added verbose function.
- Function prepare() split into convOptions() and convData().
- Numerous small changes all over the code.


Version 1.0 (11-Jun-2004)
-------------------------

- Initial version based on LIBSVM 2.6. It was written during the
  "Projektseminar Mechatronik 2004" at the Institute of Automatic
  Control, Darmstadt University of Technology, Germany.