DistributedR
DistributedR copied to clipboard
Support sparse darrays in hpdglm
The main goal should be making the package more memory efficient. The biggest darray is the one contains predictors. In some applications, the matrix of predictors are sparse so using sparse darrays may have a big impact on both memory usage and total computation time (because of reducing communication overhead). Efficient support of sparse darray for predictors will touch almost all the foreach loops of the package.