Yes, thank you. I did read it. It has an example for how to use LAPACK for QR, which is exactly what I need. I will try it out soon.
Most of the names of the LAPACK routines are pretty cryptic and a person needs to know the naming conventions. Some programs make automatic selections which LAPACK routine to call dependent on an analysis of the data and the problem on hand. I think the J package does not seem to do that. I will report back if I could get similar results from J LAPACK as from Octave or Python. Thanks. Yours sincerely, Imre Patyi On Thu, Apr 29, 2021 at 3:01 PM Gilles Kirouac <[email protected]> wrote: > >>>> Could you please tell me what I would need to change to fix the > >>>> following little J program (e.g., how to replace %. with a call > >>>> to LAPACK)? > > You should read > > https://code.jsoftware.com/wiki/Vocabulary/LAPACK > > ~ Gilles > > Le 2021-04-29 à 14:29, Henry Rich a écrit : > > %. does not use SVD. It does do QR decomposition followed by inverse. > > > > You can call LAPACK to perform the SVD, but you might need to delete > > columns with low singular values. > > > > Which routine you need to use depends on details of what you are going > > to do. > > > > Henry Rich > > > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm > ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
