math - fitness function and Selection for a Genetic Algorithm -
math - fitness function and Selection for a Genetic Algorithm -
i'm trying design nonlinear fitness function maximize variable , minimize variable b. issue maximizing much more of import @ single digit values, logarithmic. b needs minimized , in contrast a, becomes less of import when little (less one) , more of import when it's larger (>1), exponential decay.
the main goal optimize a, guess analog a=profits, b=costs
should aim maintain positive can utilize roulette wheel selection, or improve utilize rank/torunament kind of system? purpose of algorithm shape optimization.
thanks
when considering multi-objective problem goal usually identify solutions lie on pareto curve - pareto optimal set. have look here 2-dimensional visual example. when algorithm completes want set of solutions not dominated other solution. you hence need define pareto ranking mechanism take business relationship both objectives - more in depth explanation, links more reading, go here
with in mind, in order explore solutions along pareto front end you not want implementation encourages premature convergence, otherwise algorithm explore search space in 1 specific area of pareto curve. implement selection operator keeps members of each iteration's optimal set of solutions, solutions not dominated + plus parameter controlled percentage of other solutions. way encourage exploration along pareto curve.
you need ensure mutation , crossover operators tuned correctly too. novel application of evolutionary algorithms, part of problem trying identify optimal parameter set problem domain... gets interesting!!
math genetic-algorithm evolutionary-algorithm
Comments
Post a Comment