31 Oct
2018
31 Oct
'18
10:13 a.m.
Hi Gregory
I guess, although we don't recommend it. You could use expert opts to do this (increase the tolerance in the mri_ca_register step for example would decrease it). Many of the steps take a -tol <tolerance> switch that can be used to terminate the numerical minimization earlier.
You can alwo run with -openmp <nthreads> on multicore machines and get pretty significant speed increases without sacrificing accuracy. cheers Bruce
On Wed, 31 Oct 2018, Gregory Book wrote:
External Email - Use Caution
I'm curious if there are options to decrease computation time, even at the expense of quality?