Pretty useful if you want to do run classification on huge data sets. I didn't know there was a project (Apache Hama) that could do parallel (MapReduce) matrix computations - this would make it easy to do PageRank kind of calculation.