Currently the eigs method of the LinearOperator class differentiates between the CPU and GPU scenarios, where the former uses scipy routines and the latter our own implementation of a power method.
Since cupy has recently added similar routines to those of scipy, we want to add them into the backend module and have a transparent use of the same routines as done in most other cases in the library
Currently the
eigsmethod of theLinearOperatorclass differentiates between the CPU and GPU scenarios, where the former uses scipy routines and the latter our own implementation of a power method.Since
cupyhas recently added similar routines to those of scipy, we want to add them into the backend module and have a transparent use of thesameroutines as done in most other cases in the library