Hi,
Thanks for publishing this library! I gave it a try in a project of mine, and while it works great, I was wondering if it would make sense to expose the Eigen Matrix type to callers, ie. make it a template typename.
I'm using least squares to solve an objective in which the Jacobian is sparse (mostly zeros), and Eigen has SparseMatrix implementations for decompositions and solvers (see eg. SparseQR), which might be more efficient/fast than the default dense matrix ones. I must admit I haven't benchmarked performance, but figured you should be aware.
Thanks.
Hi,
Thanks for publishing this library! I gave it a try in a project of mine, and while it works great, I was wondering if it would make sense to expose the Eigen
Matrixtype to callers, ie. make it a template typename.I'm using least squares to solve an objective in which the Jacobian is sparse (mostly zeros), and Eigen has
SparseMatriximplementations for decompositions and solvers (see eg.SparseQR), which might be more efficient/fast than the default dense matrix ones. I must admit I haven't benchmarked performance, but figured you should be aware.Thanks.