The netgen package offers several methods to generate random graphs
respectively networks for benchmarking purposes. You may want to
generate some random graphs or clustered graphs to perform benchmark
studies on some fine algorithms for the travelling salesperson problem
as an example. Examples of generated networks are depicted in the
following figure. 
The package has recently been submitted to CRAN. Find additional information on the official CRAN website. To install the current developement version of the package, install the devtools package by Hadley Wickham, load it and type the following line to a R session:
install_github("jakobbossek/netgen")To generate a random network in the euclidean plane [0,1] x [0,1] we
simply load the package and call the generateRandomNetwork function,
providing the desired number of nodes.
library(netgen)
set.seed(123)
x = generateRandomNetwork(n.points = 50L)
print(x)
autoplot(x)To generate a network with points distributed around several clusters in
the [0,100] x [0, 100] space make use of the more powerful
generateClusteredNetwork function. It is mandatory to provide the
number of points n.points and the number of clusters n.cluster.
Note: See the documentation of the function to read about how it works internally.
library(netgen)
set.seed(1)
x = generateClusteredNetwork(n.cluster = 5L, n.points = 150L, upper = 100)
print(x)
autoplot(x)- Default number of digits for exportToTSPlibFormat is 10 now
- Added function isEuclidean
- rescaleNetwork now supports instances with depots
- rescaleNetwork now recomputes distance matrix
- rescaleNetwork throws error on non-euclidean networks
- Added function getTSPInstancesOverview to get an overview of files in TSPlib format
- Added function filterTSPInstances which allows to subset/filter TSP instances in TSPlib format by properties.
- Networks have an additional property edge.weight.type now (see TSPlib documentation for details)
- Added function getValidEdgeWeightTypes
- last line of TSPlib export now containts EOF string
- function importFromTSPlibFormat now has an additional argument round.distances with default value TRUE which decides whether distances of EUC_2D instances should be rounded to the nearest integer value
- added rescaleNetwork
- network objects now store the distance matrix internally
- network objects now have name and comment field
- lower and upper bound are now optional for makeNetwork/makeClusteredNetwork
- added possibility to pass point matching or point matching algorithm to morphInstances
- getOptimalPointAssignment now has another parameter method (new method to solve the assigment to optimality)
- importFromTSPlibFormat and exportToTSPlibFormat for storing networks in the TSPlib format
- importFromFile and exportToFile for storing networks in proprietary format (similar to TSPlib format, but with additional information like cluster membership or arrival/request times)
- generate{Random,Clustered,Grid}Network now automatically generate a random name if none is provided
- there is no makeClusteredNetwork function anymore. Instead makeNetwork has an optional paramter membership
- First submission to CRAN.
Please address questions and missing features about the netgen package to the author Jakob Bossek [email protected]. Found some nasty bugs? Please use the issue tracker for this. Pay attention to explain the problem as good as possible. At its best you provide an example, so I can reproduce your problem.