I am trying to use the genetic feature selection feature of KNIME (with the Feature Selection Loop nodes from the Analytics / Mining / Feature Selection node group), but I have some issues with the exact meaning of the different parameters (and their affects on the feature selection process).
My first issue is related to the early stopping.
I thought that without turning on the early stopping, the number of iterations will be ‘population size * (max number of generations + 1)’, but the loop will exit much much earlier than that (depending on the parameters, the exact number of iterations are changing).
So what is the exit criteria exactly?
Hi @kormoczi and sorry for the late reply.
I believe that if the genetic algorithm converges into a local optimum the number of iterations can be less than the maximum even if early stopping is not enable.
Is there a reason you want to force max iterations with this algorithm? My understanding is that with genetic feature selection you usually want to enable early stopping anyway.