Q: Which SVM formulation does SVM-QP and SVMPath solve?

A: SVM-QP and SVMPath solve the 2-norm soft margin SVM classification problem, exactly the same formulation as is solved by SVMlight.

Q: What is the difference between SVM-QP and SVMPath?

A: SVM-QP is a Fortran implementation of a QP solver to solve one instance of a SVM problem. The implementation contains interior point and active set algorithms. SVMPath is a C++ implementation of the active set method in SVM-QP, which is also extended to produce the entire regularization path of solution for a given range of regularization/penalty parameter values.

Q: How can I use SVM-QP and SVMPath?

A: You can compile SVM-QP into a library and call it as a subroutine by passing it the data and the labels arranged into appropriate data structures. You also can set the kernel and other parameters for the problem. For details refer to the README file distributed with the source. SVMPath can be used as a callable library or as a stand alone code. It read the data from the input file in DOC format - same format as is used by SVMlight. Various parameters for SVMPath are set in a parameter file.

Q: How big are the problems that SVM_QP/SVMPath can handle?

A: This depends on the version that you use, the available memory and the size of the optimal active set.

In Linux on and IBM (not-so-high-end) laptop we were able to solve the adult and web problems from the

UCI repository in a matter of minutes or even seconds. However, the number of optimal active support vectors (the examples that are exactly on the margin) did not exceed 1500 in these tests. If the number of active support

vectors is very large and is similar to the number of data points then SVM-QP will probably be inefficient and will run into memory problems. However, we believe that such cases result in the overfitting of the data and, hence, it is questionable whether they should ever be solved. SVMPath has very similar runtime to SVM-QP, but may be slightly slower due to C++. Also it may suffer from slowdown when there are too many breakpoints on the regularization path.

Q: I would like to try SVM-QP/SVMPath, but I don't want to spend too much time setting it up.

A: Setting up may be easier that is appears from the first glance. Please contact the project manager for help. If you can discuss your specific application we may help you estimate whether SVM-QP/SVMPath is the ideal solver for it. In the future a Matlab interface is in the plan, please check back.

Q: What kind of problem does SINCO solve?

A: SINCO (Sparse INverse COvariance selection) solves the same problem as is solved by COVSEL and Glasso. In produces a sparse positive definite matrix which is an approximation of the inverse of the covariance matrix of a multivariate Gaussian model.

Q: How can I use SINCO?

A: SINCO has a Matlab interface (provided) which make it very easy to use in that setting. It can also be used as a callable C++ library.

Q: How good is SINCO?

A: SINCO is very much under development and testing at the moment. The overall performance is being evaluated. Please check with the project manager if you want to use SINCO.

Last modified 13 years ago Last modified on Feb 17, 2009 1:42:14 PM