topic Which function could solve linear least square with inequality constraints? in Intel® oneAPI Math Kernel Library

Which function could solve linear least square with inequality constraints?

Guo__Qiandong — Thu, 14 Mar 2019 15:42:01 GMT

I want to solve X in a linear system A*X=B with a constraint that all elements in X are non-negative. It is a Non-negative Least Squares (NNLS) problem. Do you know which function in the MKL library for this problem? I tried to use trnlspbc_solve(), the Nonlinear Least Squares Problem with Linear Constraints, but it doesn't work well. Any ideas?

You may choose a suitable

mecej4 — Fri, 15 Mar 2019 04:40:15 GMT

You may choose a suitable NNLS solver from the list provided at http://plato.asu.edu/sub/nonlsq.html , under the heading "THE CONSTRAINED LSQ-PROBLEM".

Thank you for your reply,

Guo__Qiandong — Sat, 16 Mar 2019 00:23:37 GMT

Thank you for your reply, mecej4! I know "Lawson, C.L. and R.J. Hanson, Solving Least Squares Problems, Prentice-Hall, 1974" could solve it. But I prefer to use MKL to do it. Does MKL have any solutions? When I use trnlspbc_solve(), it is time-consuming and the result is a little bit different than NNLS. Any comments about using trnlspbc_solve() to do linear least square problem with constraints?

At this point it would be

mecej4 — Sat, 16 Mar 2019 05:54:13 GMT

At this point it would be useful to have a working example source code that exhibits slow performance with trnlspbc_solve(). Even a mathematical problem definition would help. I have used trnlsp() a number of times, and it has always performed quite well. Your problem may use specific functions and bounds that trigger a slowdown, which is why a full description is needed.