Inline. Sent from my iPhone
> On Mar 12, 2014, at 18:14, Ray Melton <[email protected]> wrote: > >> On 03/12/2014 05:44 PM, Ted Dunning wrote: >> I have been working with a company named 0xdata to help them contribute > ... >> *Required Additional Work* >> >> Sparse matrices >> Linear algebra bindings >> Class-file magic to allow off-the-cuff function definitions > > What kind of additional work is required in terms of sparse matrices? Congrats on reading to the end! The issue is that there needs to be a new compressed format for vectors that is efficient for sparsity less than about 1-10%. Cliff has some good approaches in mind. A less stringent problem has to do with the number of rows or columns but a quick test indicated that the practical limits are considerably higher than suspected. Many applications will be feasible with the current architecture with only a new vector compressor. > > By that, I mean did you have a sparse matrix package that just didn't working > out quite so well? Or do you have a sparse matrix package that could be a > much better sparse matrix package? Mahout has a pretty good sparse matrix package which is difficult to parallelize well. H2o has the capability to efficiently handle matrices with very limited degrees of sparsity. What we want is parallel very efficient sparse matrices. > > Ray.
