RVowpalWabbit connects GNU R to the very impressive Vowpal Wabbit fast out-of-core learning system.


As of right now (in September 2011), the R wrapping is bare-bones. The package builds cleanly, and is on the CRAN network at this page.

It should build on all Unix-alike systems; the sole external requirement (besides GNU R and the Rcpp interface between C++ and R) is the Boost.Program_options library which configure will attempt to locate.

At present, arguments are supplied in essentially the same way as for the command-line version vw, but directly from R as a vector of strings. Running demo(vw) after loading the program shows a few examples which correspond to the regression tests in the upstream sources.


From this machine, you can get the most recent tar archive, or simply peruse the entire directory.

Alternatively, you can also get it the Comprehensive R Archive Network (CRAN) node in the src/contrib directory or via at its page.


The easiest way is probably to simply say install.package("RVowpalWabbit") from inside GNU R. Alternative, R CMD INSTALL RVowpalWabbit_*.tar.gz will install from command prompt.


The initial R-Forge repo, has since been moved to a new GitHub repository.


RVowpalWabbit is written by Dirk Eddelbuettel.


RVowpalWabbit is licensed under the GNU GPL (>= 2(

Last modified: Tue Jan 7 21:38:22 CST 2014