I know using sparse matrices can result in drastic speed enhancements. I know any matrix is convertible to a sparse matrix. From the top-down, it seems like OpenCV and libsvm would use this, but there's a data type called [SparseMat]. I've looked at the libsvm FAQ because my specific use requires SVM; it says that they only support the sparse format.
Does OpenCV and libsvm assume a matrix is sparse or convert it to a sparse matrix? Should I convert my matrix to a sparse matrix before I use OpenCV's SVM?