The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.
The package noisyCE2
implements the cross-entropy
algorithm (Rubinstein and Kroese, 2004) for the optimisation of
unconstrained deterministic and noisy functions through a highly
flexible and customisable function which allows user to define custom
variable domains, sampling distributions, updating and smoothing rules,
and stopping criteria. Several built-in methods and settings make the
package very easy-to-use under standard optimisation problems.
The negative 4-dimensional paraboloid can be maximised as follows:
<- function(x) { -sum((x - (1:4))^2) }
negparaboloid
<- noisyCE2(negparaboloid, domain = rep('real', 4)) sol
The 10-dimensional Rosenbrock’s function can be minimised as follows:
<- function(x) {
rosenbrock sum(100 * (tail(x, -1) - head(x, -1)^2)^2 + (head(x, -1) - 1)^2)
}
<- type_real(
newvar init = c(0, 2),
smooth = list(
quote(smooth_lin(x, xt, 1)),
quote(smooth_dec(x, xt, 0.7, 5))
)
)
<- noisyCE2(
sol domain = rep(list(newvar), 10),
rosenbrock, maximise = FALSE, N = 2000, maxiter = 10000
)
The negative 4-dimensional paraboloid with additive Gaussian noise can be maximised as follows:
<- function(x) { -sum((x - (1:4))^2) + rnorm(1) }
noisyparaboloid
<- noisyCE2(noisyparaboloid, domain = rep('real', 4), stoprule = geweke(x)) sol
where the stopping criterion based on the Geweke’s test has been adopted according to Bee et al. (2017).
Bee M., G. Espa, D. Giuliani, F. Santi (2017) “A cross-entropy approach to the estimation of generalised linear multilevel models”, Journal of Computational and Graphical Statistics, 26 (3), pp. 695-708. https://doi.org/10.1080/10618600.2016.1278003
Rubinstein, R. Y., and Kroese, D. P. (2004), The Cross-Entropy Method, Springer, New York. ISBN: 978-1-4419-1940-3
These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.