numpy - How can I parallelize functions "leastsq" or/and "curve_fit" -
what best way parallelize fitting procedure multicore computers using scipy functions? far see manual, these functions not have parameters npocs. mean not supposed parallelized?
short answer: there no built-in parallelization @ moment. there a proposal utilize multiprocessing in leastsq, it's not clear whether worth it. (if sense giving code drive , study results, that'll appreciated).
that beingness said, can parallelize evaluation of objective function wish. eg force compiled extension (cython, c, fortran), release gil , utilize openmp or explicit threads. mileage may vary a lot depending on details of particular problem.
numpy parallel-processing scipy mathematical-optimization curve-fitting
No comments:
Post a Comment