Techniques of evolutionary algorithms (EA) rely heavily on the use of random number generators (RNGs). From initial population generation, through the specific canonical operators applied to create new temporary population, the use of randomness is pervasive through EAs.

Best quality is believed to stem from RNGs that produce number sequences that are uniform and uncorrelated.

Two different RNGs were applied to an EA, which calculated values for populations over time. The goal was to find out how quality of RNGs affect EAs performance.

How can the sensitiveness of the EA to the two RNGs be detected?

  • $\begingroup$ Do you have any specific reason to think this is a problem with evolutionary algorithms? did you see any bad examples? $\endgroup$ – kjetil b halvorsen Jul 10 '18 at 21:45

With the good, modern RNG's now available (like those implemented in R), this is probably not a very important or big problem. However, I suppose there is some meta-theorem saying that for any RNG, there is some (maybe very artificial) problem where the structure of the RNG and the structure of the problem will interfere. Just try your algorithm with different RNG's and look if something suspicious arises. Or just ignore the "problem" until you have some evidence the solutions proposed by your algorithm is bad!

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.