Theoretically yes, when you do not set PARAMETERS, but FACTORS. In train mode it then loads the previously trained parameters and trains OptimalF. This is not documented, so try it at your own risk.