Target handling for the optimisation algorithm

In GitLab by @hehnen1 on Nov 22, 2017, 19:16

In cases where one has more then one repetition of an experiment, the question raises on how to treat all the experimental data as the target for the optimisation algorithm. How are the results to be processed to become a meanigful target. Would a mean value per time step be sufficient or rather a kind of bonus system within an area range that covers all experiments and gives the highest bonus.