[R] Noisy objective functions

J C Nash pro|jcn@@h @end|ng |rom gm@||@com
Mon Aug 14 02:04:07 CEST 2023

More to provide another perspective, I'll give the citation of some work
with Harry Joe and myself from over 2 decades ago.

   author  = {Joe, Harry and Nash, John C.},
   title   = {Numerical optimization and surface estimation with imprecise function evaluations},
   journal = {Statistics and Computing},
   year    = {2003},
   volume  = {13},
   pages   = {277--286},

Essentially this fits a quadratic approximately by regression, assuming the returned
objective is imprecise. It is NOT good for high dimension, of course, and is bedeviled
by needing to have some idea of the scale of the imprecision i.e., the noise
amplitude. However, it does work for some applications. Harry had some success with
Monte Carlo evaluation of multidimensional integrals optimizing crude quadratures.
That is, multiple crude quadrature could be more efficient than single precise quadrature.
However, this approach is not one that can be blindly applied. There are all sorts
of issues about what point cloud to keep as the "fit model, move to model minimum,
add points, delete points" process evolves.


On 2023-08-13 15:28, Hans W wrote:
> While working on 'random walk' applications, I got interested in
> optimizing noisy objective functions. As an (artificial) example, the
> following is the Rosenbrock function, where Gaussian noise of standard
> deviation `sd = 0.01` is added to the function value.
>      fn <- function(x)
>                (1+rnorm(1, sd=0.01)) * adagio::fnRosenbrock(x)
> To smooth out the noise, define another function `fnk(x, k = 1)` that
> calls `fn` k times and returns the mean value of those k function
> applications.
>      fnk <- function(x, k = 1) {     # fnk(x) same as fn(x)
>          rv = 0.0
>          for (i in 1:k) rv <- rv + fn(x)
>          return(rv/n)
>      }
> When we apply several optimization solvers to this noisy and smoothed
> noise functions we get for instance the following results:
> (Starting point is always `rep(0.1, 5)`, maximal number of iterations 5000,
>   relative tolerance 1e-12, and the optimization is successful if the
> function value at the minimum is below 1e-06.)
>        k   nmk       anms neldermead     ucminf optim_BFGS
>       ---------------------------------------------------
>        1  0.21       0.32       0.13       0.00       0.00
>        3  0.52       0.63       0.50       0.00       0.00
>       10  0.81       0.91       0.87       0.00       0.00
> Solvers: nmk = dfoptim::nmk, anms = pracma::anms [both Nelder-Mead codes]
>           neldermead = nloptr::neldermead,
>           ucminf = ucminf::ucminf, optim_BFGS = optim with method "BFGS"
> Read the table as follows: `nmk` will be successful in 21% of the
> trials, while for example `optim` will never come close to the true
> minimum.
> I think it is reasonable to assume that gradient-based methods do
> poorly with noisy objectives, though I did not expect to see them fail
> so clearly. On the other hand, Nelder-Mead implementations do quite
> well if there is not too much noise.
> In real-world applications, it will often not be possible to do the
> same measurement several times. That is, we will then have to live
> with `k = 1`. In my applications with long 'random walks', doing the
> calculations several times in a row will really need some time.
> QUESTION: What could be other approaches to minimize noisy functions?
> I looked through some "Stochastic Programming" tutorials and did not
> find them very helpful in this situation. Of course, I might have
> looked into these works too superficially.
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

More information about the R-help mailing list