Hi, all. I have been testing out these routines to compute linear regressions with dense matrices https://eigen.tuxfamily.org/dox/group__ ... uares.html
and there is also awesome routines that allow for the solution of A*x=b with matrix-free format http://eigen.tuxfamily.org/dox/group__M ... ample.html.

But weirdly, I could not use any of the least-squares routines with matrix-free class that implements matrix-vector multiplications. In my case I cannot afford to instantiate A matrix, but I can code up the functions that multiply it by a vector or a matrix.

There is a similar package in Julia Statistics package called lsmr, that works with Linear Operators, which are basically, the Julia implementation of Eigen matrix-free class.

Am I missing something or is matrix free linear regression impossible in Eigen?