linfa
linfa copied to clipboard
Support sparse matrices
In high dimensional statistics, we usually rely on CSC and CSR arrays when matrices are very sparse. It is particularly for speeding up the coordinate descent procedure.
I can implement that in the ElasticNet subpackage if you find it worth it. What do you think?
The sprs crate implements CSR matrices and is already used in linfa-kernel and linfa-processing. Integrating it into ElasticNet would be welcome.