jluby / walkr

This package implements MCMC random walks in the intersection of the n-simplex and hyperplanes

walkr

Consider the intersection of two spaces: the complete solution space to Ax = b and the N-Simplex. The intersection of these two spaces is a convex polytope. walkr samples from this intersection using two Monte-Carlo Markov Chain (MCMC) methods: hit-and-run and Dikin walk. walkr also provide tools to examine sample quality.

Getting Started

• Install from CRAN(released version):

  install.packages("walkr")

• Install from GitHub (development version):

  devtools::install_github("andyyao95/walkr")

Sampling Points

  library(walkr)  
  A <- matrix(1, ncol = 3)  
  b <- 1    
  sampled_points <- walkr(A = A, b = b, points = 1000, 
                          method = "dikin", ret.format = "list")   

Visualizing the Sampled Points

explore_walkr(sampled_points)