Linear least squares solver with a constraint

Open discussion for MAGMA library (Matrix Algebra on GPU and Multicore Architectures)

Linear least squares solver with a constraint

Postby Robert17 » Tue Nov 07, 2017 3:40 pm

Does the MAGMA library support a linear least squares solver subject to an inequality? I'm trying to write a CUDA equivalent of a MATLAB function called lsqlin.
https://www.mathworks.com/help/optim/ug ... tid=doc_ta
Anyone know how I can do this?
Robert17
 
Posts: 2
Joined: Tue Nov 07, 2017 3:30 pm

Re: Linear least squares solver with a constraint

Postby mgates3 » Wed Nov 08, 2017 11:55 am

It sounds like you want the equivalent of the LAPACK routine gglse (least squares with equality constraints).

http://www.netlib.org/lapack/double/dgglse.f

We don't currently have that functionality in MAGMA. We only have the regular least squares (gels).

-mark
mgates3
 
Posts: 750
Joined: Fri Jan 06, 2012 2:13 pm

Re: Linear least squares solver with a constraint

Postby Robert17 » Wed Nov 08, 2017 12:55 pm

Yes, we want a CUDA version of the LAPACK routine. Thanks for letting me know. I'm currently talking to the developers to see if they can add this routine.
Robert17
 
Posts: 2
Joined: Tue Nov 07, 2017 3:30 pm


Return to User discussion

Who is online

Users browsing this forum: Bing [Bot] and 5 guests

cron