[R] "Heuristic optimisation"?

[Ben Bolker]

Ajay Shah <ajayshah <at> mayin.org> writes:

> 
> I wondered was people on this list felt about this article:
>   http://www.voxeu.org/index.php?q=node/2363
> which talks about the problems of obtaining sound answers in numerical
> optimisation in settings such as MLE or NLS.

  It seems perfectly reasonable (it says that lots of real-world
optimization problems have multiple maxima and rough surfaces 
and that a variety of stochastic global optimization techniques
like simulated annealing, genetical algorithms, differential
evolution, etc., are useful) but nothing desperately new.
R does have some tools (genoud package, method="SANN" in
optim) for heuristic optimization etc. but (perhaps because
they generally require a lot of tuning/fiddling) they
aren't as generally available and polished as the optimization methods that
assume a smooth/unimodal surface ...

  Ben Bolker