read.svmlight() and write.svmlight() read/write sparse matrices in SVMlight format.
You will find other functions for this on the internet, but the ones I found were either slow or handled only dense (=normal) matrices.
Benchmark to convert a dgCMatrix with 2,500,000 rows and 8,500 columns (1.1GB in memory) => 5 minutes
You would take hours if not days with regular packages for such sizes!
devtools:::install_github("Laurae2/sparsity")Original source (felixr/sparsity) does not compile properly! Hence the reason of this repository for those who needs only the SVMLight format >_>