Print

Print


ECJers,

I am working with a 3-objective problem where some objectives are to be maximized and some are to be minimized with NSGA-II and SPEA2.  However I set my parameters, however, all three appear to be maximized.

I have tried this:

multi.fitness.0.maximize=true
multi.fitness.1.maximize=false
multi.fitness.2.maximize=false

And this:

pop.subpop.0.species.fitness.0.maximize=true
pop.subpop.0.species.fitness.1.maximize=false
pop.subpop.0.species.fitness.2.maximize=false

The result still shows pairs like this in the Pareto front, which shows that objective 1 is being maximized:

Objective 0 Objective 1 Objective 2
0.2715239 0.28555658 4.0
403 0.2809572 0.26017603 4.0

I'd like to know if I should hunt for a bug in ECJ, or if I'm setting the parameters wrong.  Uday told me he observed the same problem some time ago.

For now I'm just converting all my objectives to maximization problems manually as a workaround (i.e. by taking the reciprocal of the true fitness).

Siggy

--

Ph.D student in Computer Science
George Mason University
http://mason.gmu.edu/~escott8/