The system includes several specialised matrix types:
- hermitian
- square
- symmetric
- triangular
There are also existing versions of these matrix types in numerical-utilities. IMO, the implementation in numerical-utilities far outclass those in linear-algebra. The only thing missing from them is a lisp-based solver, which linear-algebra does have. There are LAPACK solvers in LLA, for those specialised matrices though.
This issue is to track a review of both of these and (probably) figure out a way to implement the solve function on the matrix implementations in numerical-utilities. An alternative may be to say that we're not going to implement a lisp based solver. I have used the gauss solver in an implementation of lowess, and find the speed very slow. This might be the reason that a lisp-based solver was never implemented for the specialised types in numerical-utilities.
The system includes several specialised matrix types:
There are also existing versions of these matrix types in numerical-utilities. IMO, the implementation in numerical-utilities far outclass those in
linear-algebra. The only thing missing from them is a lisp-based solver, whichlinear-algebradoes have. There are LAPACK solvers in LLA, for those specialised matrices though.This issue is to track a review of both of these and (probably) figure out a way to implement the
solvefunction on the matrix implementations innumerical-utilities. An alternative may be to say that we're not going to implement a lisp based solver. I have used the gauss solver in an implementation of lowess, and find the speed very slow. This might be the reason that a lisp-based solver was never implemented for the specialised types innumerical-utilities.