5
$\begingroup$

Genetic algorithms are avoided in econometry literature as often as possible, but still sometimes they are inevitable. The question is: Which well known models are the most difficult to estimate using conventional algorithms? (By Conventional algorithms I mean Gauss-Newton method, Levenberg–Marquardt algorithm, and so on)

Motivation: I want to test some heuristic methods, and I need some benchmark to be sure that this particular model is really hard to estimate.

$\endgroup$
1
  • 1
    $\begingroup$ I'd say any model that involves lots of sines/cosines or decaying exponentials would make for badly-behaved problems. $\endgroup$ – J. M. is not a statistician Dec 17 '10 at 5:27
3
$\begingroup$

The travelling salesman problem is surely an archetypal hard optimization problem. To quote Wikipedia, "it is used as a benchmark for many optimization methods".

$\endgroup$
3
  • $\begingroup$ I know many problems not related to econometry which are used as a benchmark, but I want to focus on applications of GA in economics, and I thought that estimation of econometry model is good start. $\endgroup$ – Tomek Tarczynski Nov 16 '10 at 22:37
  • $\begingroup$ Surely the travelling salesman problem has huge application within economics. Perhaps you could clarify what you mean by an 'econometry model'. $\endgroup$ – onestop Nov 17 '10 at 8:09
  • $\begingroup$ Sorry it was late, of course I meant econometric model. Example: Y=b1*x1+b2*x2^b3+(b4/x3+b5*x4)^b6 , but I'm not sure if this is hard to estimate using non-linear regression and secondly whether it has any real-world application. $\endgroup$ – Tomek Tarczynski Nov 17 '10 at 9:36

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.