I looked at them a while ago, and part of the issue is that general-purpose vectors need to be able to do all things for all people... when you give that up, and say that your sparse vectors are, say, immutable, after creation, and only designed to do operations like dot(DenseVector) and DenseVector.plus(SparseVector) (and maybe DenseVector.plus(SparseVector, double multiple), for doing u + b*v), then you can get away with being very simple - just parallel arrays of indexes and values for the sparse vectors, and double[] arrays for the dense ones.
I'll see if I can compare performance of my specialized simple vectors and general purpose ones, and see what the difference is. -jake On Fri, Sep 25, 2009 at 12:08 PM, Ted Dunning <[email protected]> wrote: > Consider improving the versions in Mahout while you are at it! > > On Fri, Sep 25, 2009 at 12:03 PM, Jake Mannix <[email protected]> > wrote: > > > but I'm currently using mostly my own sparse and dense > > vector writables for use on Hadoop (designed specifically for things like > > Lanczos and AGHA), so I'd need to port them over to use whichever vector > > impls Mahout is using > > > > > > -- > Ted Dunning, CTO > DeepDyve >
