Particle swarm UseParallel issur

I use particle swarm to fit a model to my data by minimizing an error function. When I run it serially it gives me the right answer but when I try to use the parallel option it gives me a nonsense fit and a value of 0 for the objective function. Has anybody encountered a similar problem?

Answer 1

[Walter Roberson]

That can happen if the fit function depends upon global variables, or upon persistent variables that are established beforehand . Or in some cases involving "objects" that are not serializable.