"Hit and Run" sampler
This project provides an implementation of the "Hit and Run" algorithm for sampling from convex shapes. Our motivation for doing this is to enable the fast generation of arbitrarily constrained weights in high-dimensional space. This should enable applying e.g. SMAA (http://smaa.fi/) with more complex weight constraints.
Releases of the R package are available on CRAN:
The source is available from GitHub:
See ??har, ?har, ?har-constraint and ?simplex.sample
hitandrun depends on rcdd, which requires GNU MP. On debian derivatives, the compile-time dependency is provided by libgmp-dev.
Thanks to Tommi Tervonen, Douwe Postmus, Ahmet Kandakoglu and Alexander Karlsson for reporting bugs and/or suggesting improvements.