DTRNLSP_SOLVE won't stop for a high valued Jacobian

andimb — Thu, 28 Feb 2008 16:46:25 GMT

Hi,

if during a nonlinear least squares problem run the Jacobi matrix is complitly filled with very high values (near infinity like 1.0E+272), a run of function DTRNLSP_SOLVE won't stop. How could I prevent this problem and why does it happen? Shouldn't the function stop after a certain kind of iterations or if a new request is needed?

Thanks

Re: DTRNLSP_SOLVE won't stop for a high valued Jacobian

Sergey_G_Intel — Mon, 10 Nov 2008 03:30:53 GMT

Hi,

First of all, thank you for using Intel MKL!

Current implementation of MKL non-linear solvers does not control the values of the Jacobian. However, you can set the maximal number of iterations for the solver and it will stop as soon as it reaches this value (so, it will not proceed with computations forever). If you found that the solver goes beyond the maximal number of iterations, please send us a test case. You may also send us your problem as a test case and we will check if everything is OK in your code with respect to the usage of the solver.

Thank you,

Sergey G

topic DTRNLSP_SOLVE won't stop for a high valued Jacobian in Intel® oneAPI Math Kernel Library

DTRNLSP_SOLVE won't stop for a high valued Jacobian

Re: DTRNLSP_SOLVE won't stop for a high valued Jacobian