Supplementary Code for the Paper:

Common Failure Modes of Subcluster-based Sampling in Dirichlet Process Gaussian Mixture Models - and a Deep-learning Solution

This repository contains the code to train the SplitNet models and run in within the DPMMSubCluster.jl package.

An example notebook with some results (Examples.ipynb) is available.
To install the run the DPMMSubCluster.jl package, refer to it's documentation.
To train your own SplitNet models, read the Readme file in the splitnet directory.
To run Sub-* methods with your own trained models, please update path to the trained models in the config file at DPMMSubClusters.jl/configs/basic.yaml