This repository contains R-code for running a simulation study on a Tardis Cluster and was developed as part of my work as a student assistant for Andreas Brandmaier in cooperation with Aaron Peikert at MPIB (Max-Planck Institute of Human Development) in Berlin, see https://github.com/aaronpeikert/maketest for the general idea developed by Aaron Peikert.
bnnlib
is a R-package (under development as of Sept, 2020) by Andreas Brandmaier that allows the flexible creation of different neural network models in R. The above code requires a version of bnnlib
in the project folder. See how to install bnnlib
here: https://github.com/brandmaier/bnnlib
Update: bnnlib
can now be installed and loaded via install.packages("bnnlib")
and library(bnnlib)
.
One of the examples in the bnnlib
vignettes aims at predicting different frequencies e.g. of some oscillating object (see frequencies.Rmd
in bnnlib/vignettes/
).
This example was extended here to test the influence of different number of layers and different number of cells per layer on training and test errors.