Stein discrepancy experiments

This repo contains many experiments that utilize the Stein discrepancy. The experiments here use the Julia package SteinDiscrepancy.jl and implement all the experiments from the following papers:

• Measuring Sample Quality with Stein's Method
• Measuring Sample Quality with Diffusions
• Measuring Sample Quality with Kernels

All the papers above introduce and explain the significance of the Stein discrepancy. The first paper has a more gentle introduction describing how the Stein discrepancy bounds standard probability metrics like the Wasserstein distance.

So how do I run these experiments?

The code has all been tested on Julia v0.5. There is a file src/startup.jl which simply adds the paths of some modules to the LOAD_PATH variable. You should either simlink this file to ~/.juliarc.jl or add the lines from this file to your current .juliarc.jl.

There is a REQUIRE file that demarcates all the necessary packages needed to run all the experiments. All experiments should be run from the base directory of this repo. For example, the command

julia src/experiments/multimodal_gmm_langevin_floor --gap=5 --seed=10

would run the experiment in the file src/experiments/multimodal_gmm_langevin_floor.jl and pass in the parameters gap and seed (defined in the script) to be 5 and 10 respectively. All experiment results will be deposited in the directory results in the base directory.

For more information, see SteinDiscrepancy.jl.