Limiting function evaluations in gamultiobj

Hi there,

I am trying to perform a multi-objective optimisation of a function with gamultiobj to obtain a pareto front. Part of the specification of the optimisation is that no more than 50,000 function evaluations should be completed. Is there a way to limit funccount so that it is always less than 50,000? I know it is based on population size and maximum generations but I am unsure what the relationship is.

Thanks in advance!

Answer 1

[Alan Weiss]

It depends on whether you perform vectorized function evaluations ( options.UseVectorized = true ) and whether you have nonlinear constraints. With no vectorization and no nonlinear constraints, then I believe that the fitness function is evaluated once per generation per individual. When there are nonlinear constraints, the process is much more involved. When there is vectorization and no nonlinear constraints, the fitness function is evaluated once per generation, all individuals at once. See How The Genetic Algorithm Works. So with no nonlinear constraints, you can limit the function count by limiting the number of generations.

I do not know how to limit this number effectively in the presence of nonlinear constraints.

Alan Weiss

MATLAB mathematical toolbox documentation