Direct Optimization of Hyper-Parameter

February 10, 2019

(This article was first published on S+/R – Yet Another Blog in Statistical Computing, and kindly contributed to R-bloggers)

In the previous post (, it is shown how to identify the optimal hyper-parameter in a General Regression Neural Network by using the Sobol sequence and the uniform random generator respectively through the N-fold cross validation. While the Sobol sequence yields a slightly better performance, outcomes from both approaches are very similar, as shown below based upon five trials with 20 samples in each. Both approaches can be generalized from one-dimensional to multi-dimensional domains, e.g. boosting or deep learning.

net <-[, -14]), Boston[, 14], sigma = 1)
sb_out <- Reduce(rbind, Map(function(x), gen_sobol(0.1, 1.0, 20, x), 4, 2019), seq(1, 5)))

uf_out <- Reduce(rbind, Map(function(x), gen_unifm(0.1, 1.0, 20, x), 4, 2019), seq(1, 5)))

Map(function(x) x[x$R2 == max(x$R2), ], list(sobol = sb_out, uniform = uf_out))
# $sobol
#  sigma        R2
# 0.5568 0.8019342
# $uniform
#  sigma        R2
# 0.5608 0.8019327

Other than the random search, another way to locate the optimal hyper-parameter is applying general optimization routines, As shown in the demonstration below, we first need to define an objective function, e.g. grnn.optim(), to maximize the Cross-Validation R^2. In addition, depending on the optimization algorithm, upper and lower bounds of the parameter to be optimized should also be provided. Three optimization algorithms are employed in the example, including unconstrained non-linear optimization, particle swarm optimization, and Nelder–Mead simplex optimization, with all showing comparable outcomes to ones achieved by the random search.

To leave a comment for the author, please follow the link and comment on their blog: S+/R – Yet Another Blog in Statistical Computing. offers daily e-mail updates about R news and tutorials on topics such as: Data science, Big Data, R jobs, visualization (ggplot2, Boxplots, maps, animation), programming (RStudio, Sweave, LaTeX, SQL, Eclipse, git, hadoop, Web Scraping) statistics (regression, PCA, time series, trading) and more...

If you got this far, why not subscribe for updates from the site? Choose your flavor: e-mail, twitter, RSS, or facebook...

Comments are closed.

Search R-bloggers


Never miss an update!
Subscribe to R-bloggers to receive
e-mails with the latest R posts.
(You will not see this message again.)

Click here to close (This popup will not appear again)