How to know if a genetic algorithm is working? -

it important benchmark algorithm before applying specific question. otherwise, garbage in, garbage out. having implemented genetic algorithm (ga) elitism, have no idea how test whether works or not.

i have thought printing out statistics, such mean, median, , variance, of each generation. however, not strong indicators of correctness. example, maximum fitness doesn't anything, because random search elitism, have non-decreasing maximum fitness. mean , median not either, because may deteriorate, if ga correct.

what effective way of testing if ga working good?

my way of deciding if algorithm working draw plot of fitness value on time of execution. this:

from question on ga. y-axis fitness level of best individual (less better), x-axis time or number of iterations. ignore red line - not relevant question

from graph determine if algorithm stuck in rut , not improve further.

however, had advantage of knowing solution , compare solution result of ga execution. , way tell if ga enough or not. not available, can try solving simple sub-set of problem , compare result result of ga.

just note, conclusion of thesis ga can give starting point other algorithms find better solution. , ga not fit find best solution complex problem many constraints.

so can try hard great result, nature of ga not give you.